In the summer of 2024, Ukraine controlled a territory in Kursk that relied on a single logistical route from a city called Sumy (in Ukraine) to a city called Sudzha (in Russia). It was against this bottleneck that Russia began to test a drone technology that has since become a defining weapon on both sides of the war: fiber-optic tethered drones.

Ukraine and Russia have since increased adoption of fiber-optic drones because they have one advantage that traditional drones do not: they don’t rely on wireless communication.

Why? What’s the deal with wireless? Two things, which go far beyond drones.

1. Just about everything in the world relies on wireless connection.

2. The technology to jam or interrupt wireless communications has far outpaced the technology to avoid jamming.

In warfare, for instance, the mere act of transmitting wirelessly—even with encryption—is what enables a company like CX2 to track down and engage drone operators. And it’s what has driven the Russian and Ukrainian militaries, among others, to purchase fiber-optic tethered drones.

But this goes way beyond the battlefield. Any time information travels wirelessly (hospital records, self-driving cars, satellite feeds, infrastructure controls), the signal itself is a vulnerability. And, as we’ll cover, current encryption is not capable of fixing that.

Enter Rampart, the company we are covering today. Its team, which includes ex-NSA PhD scientists, mathematicians and engineers, has developed a brand new technology that makes wireless signals look the same as the background noise. If signals can’t be identified, they can’t be intercepted or jammed. This is revolutionizing the future of wireless communications.

So, what exactly is Rampart working on? What is it like to work there? Why should you consider joining them? I sat down with the team to get answers to those questions and more.

The product

Imagine it is a sunny Friday afternoon and you are sitting down at a neighborhood café with a friend. The conversation turns to a subject you aren’t comfortable speaking about in public. So you suggest to your friend that the two of you start speaking in Spanish, a language you are both fluent in. You hope nobody around you speaks Spanish.

The Spanish-speaking approach is a rough (but useful) analog for modern wireless encryption. The message gets scrambled into something that is hopefully unintelligible. But everyone around you can still see that you’re talking. They can see who you are, who you’re talking to, when the conversation started, and how long it’s lasted. If they wanted to, they could come over and disrupt the conversation. They could even follow you home.

What if you didn’t want people to know you were talking at all?

Then you’d need to figure out some sort of communication system that was undetectable. A way of speaking to each other that made your sounds just like the background noise in the café, a way of speaking that only the two of you could possibly understand. Our hypothetical now fails us, because it seems impossible. And yet it is what Rampart is building.

The core innovation behind Rampart’s products is called UBDM and a related technology known as RSW. These technologies scramble the entire wireless signal so it looks exactly like random background noise. UBDM hides what you say (and then some), and RSW hides that you are saying anything at all. In both cases, there is no chance for an eavesdropper to intercept (recover the data in the signal).

This is a helpful visual:

In other words, with Rampart, your wireless communications are pure background noise to everyone except for you and the person (or device) on the other end of the comms.

Right now, this technology plays out in two distinct products. UBDM Core is the raw innovation itself, sold as a licensable IP block. It’s a piece of logic that can be dropped into any radio platform; the engine that other people build into their own radios and systems.

The second is the StrataWave UAS (Unmanned Aircraft System) Radio. Rampart’s two products both use the UBDM and RSW in different ways. UBDM Core enhances almost any existing system, preventing interception and providing persistent and secure connectivity without needing cables. StrataWave is a drone radio engineered for military UASs operating in contested airspace; it makes drones extremely difficult to detect or jam.

Rampart’s products, to date, have been sold for military applications. Talk to anyone at the company, though, and you will learn that the vision is far more ambitious than the battlefield.

The strategy

Warfare is not the only application for Rampart’s technology. Because, as I mentioned earlier, wireless connectivity is required for most things in the world to work. Hospitals, financial institutions, satellites, power grids, water systems—these are a few other examples of areas where Rampart’s technology could have a big impact.

“Rampart is framing a new field of discovery at the intersection of Physics and Information Theory,” Matt Robinson, Rampart’s founder and Chief Scientist told me. “We are poised for total leadership from the physical layer up, for defining the frontier of communication technology, and for being the global standard for all network engineering,

In this new version of the world where standard wireless and standard encryption is not enough, you can envision Rampart being useful for all wireless communications; perhaps starting with the most sensitive (as they have with the military) and working backwards from there.

One belief driving Rampart’s approach to the business is that “we’ve largely reached the end of easy wins. The low-hanging fruit is picked over, and even most of the medium-sized problems are quickly being solved. What’s left are the ‘hard’ problems - the ones that won’t yield to patchwork solutions or half-measures.”

“If we want the next great leap in history, we have to stop chasing shortcuts and start reinvesting in the hard, slow, substantive work of fundamental engineering, science, and mathematics. Real progress requires real thought, not just harder sprints.”

“We have a fundamentally novel and powerful approach to the next generation of wireless communications,” Rampart’s VP of Growth, Dzu Do, said. “This is the culmination of years of research and development applied towards solving real world problems in both the defense and commercial industries.”

The growth

If the tech startup world is a stereotypical high school classroom, Rampart is not one of the kids raising their hand every two seconds. They are not one of the kids playing video games in the corner. They are not one of the kids scrolling Instagram or debating politics with the professor. No, Rampart feels a little more like the nerdy smart kid sitting quietly in the back, taking notes.

The company doesn’t try much to market to the general public. Their official X account has 0 posts. There are no splashy Series A announcements or big VC names. Rampart has instead raised money behind the scenes through a combination of grants and funding from angel investors; a path more common among defense companies than ‘hot’ software startups.

Rampart’s distribution so far has been mostly through defense ecosystem relationships: SBIR contracts, partnerships with companies like Lockheed Martin, Comtech, and Thales Defense and Security, and direct Pentagon engagement. The company hosted former Under Secretary of Defense Heidi Shyu (the third-highest-ranking civilian in the DoD), Dr. Tom Rondeau (Principal Director for FutureG) and Alex Lovett (recently retired Director of Prototypes and Experiments for Mission Capabilities) at their facility, which is a strong sign of DoD interest in Rampart at the highest levels.

The culture and team

There is one stereotype of startup culture that focuses on MVPs, on pushing to production as fast as possible, on spending as little on development as possible. That is not Rampart.

I talked to Jay Patel, Rampart’s EVP of Product and Engineering, who previously worked on some of the first mobile phones at Motorola and led product teams at Avaya and Vonage. “We invest heavily in true research,” he said. “This is rare to see as many startups focus more on… [just] trying to get a product to market.”

The R&D-first ethos is so strong that Rampart’s CEO, Omar Javaid, thinks it might even be controversial. “We have some of the best mathematicians and physicists on the planet! It’s unpopular because it’s a heavy, long-term investment and most people think short term.”

The team “is unique. They have a deep technical background, were successful academics, and - crucially - applied their research interests at the NSA and in the field with actual operators.”

(The R&D claim isn’t just fluff; Rampart holds 130+ patents!)

You might reasonably assume that working at Rampart alongside some of the ‘best mathematicians and physicists on the planet’ is probably not an easy job. That would be correct. “It’s really challenging to work on fundamental technology where the timing is uncertain. Some people are attracted to uncertainty but find that environment difficult to actually work in.”

The biggest driver of talent to Rampart is the novelty of the technology. It is an opportunity for PhDs to work on revolutionary technology. Also, a common theme amongst all employees is the fundamental goal to keep our servicemembers safe when they are working in contested areas.

Rampart is based in Linthicum Heights, Maryland, and the team works in-person.

Should you join Rampart?

The word “inventor” often sounds like an old-fashioned thing that does not apply to the work people do today. It can feel like lots of tech startups are just building slightly better versions of things that already exist, step-changes that might be useful but are not very exciting.

Rampart is inventing things in the old-school sense of the word. They have “made advances that were previously considered impossible.” Their goal is to become the gravity well for a new era of wireless. If you are a smart person interested in working on a big problem, this probably ranks within the top decile (or higher) or projects you might want to consider.

Matt Robinson, Rampart’s founder, told me that he wants the company to be the place that people “send their best PhD students and professors on sabbatical because this is where the most exciting and groundbreaking work happens.”

The first value Rampart lists on their general job application is: “Privacy is a right.” If you are interested in building a more secure, more private version of wireless for the future, and if you would like to do that alongside some extremely smart and hardworking people, you may want to consider joining Rampart. You can see their open roles here.

Thanks to Rampart for supporting Next Play and making this Spotlight possible

Y Combinator famously offers two big pieces of advice to startups:

1. Write code.

2. Talk to users.

I’d argue that this is good practice for all companies; making a better product and understanding your users is generally upstream of earning more profit.

It also feels like we sort of have the first item on that list covered. Or at least we’re getting close. The comparatively underrated part seems to be that second piece of YC advice: talk to users.

Historically, if you needed to run a study on customer, product, or market, there was no great way to approach it. Research has long been one of the slowest, most expensive, and most unreliable processes in business.

The company we are writing about today, Listen Labs, is transforming that status quo by building a new kind of customer research platform that is faster and more cost-effective without compromising on quality.

The pitch is simple: Listen accelerates your research by designing studies, sourcing participants, conducting interviews, analyzing results, and making them easily presentable. Traditionally, this would be an expensive endeavor that takes months and requires working with multiple legacy solutions like Qualtrics, Gartner, or AlphaSights. Using a carefully trained AI researcher infused with best practices, Listen shortens the research to results loop to mere days.

Listen Labs announced a $69M Series B in January and already counts hundreds of massive enterprises as customers. So we chatted with their team to figure out how it all works: what is the plan for the business? Why do they believe it will win instead of competitors? What makes the culture unique? And should you join?

The product

Market research has been stagnant for decades. “Qualtrics has been running the same legacy-style surveys for years and is still worth $12B,” Alfred Wahlforss, CEO of Listen Labs, told me when asked about the market. “A single Fortune 500 company might spend $10M on Qualtrics and $100M on research agencies.”

Usually, “even on an expedited timeline you’re talking 4 to 6 weeks from true question to insight and action,” said Brian Davia, Head of Consumer and Business Insights at Sweetgreen, a Listen customer.

With the painstaking manual work done with AI, Listen Labs says you can go from a question to a useful answer in as fast as a few hours.

Here is one example of how you might interact with the product. Say you want to know which of two ad concepts resonates better with Gen Z women in tech. You tell that to Listen and their AI turns your plain English question into an effective interview plan that you review and approve. Then Listen sources young women in tech from its pool of over 30M people (or, your own user database if you prefer) and dispatches an AI interviewer to talk to them, simultaneously, in real-time video or audio conversations. Just like a human would, with thoughtful follow-up questions and clarifications.

A few hours later, you have an in-depth analysis of the findings, all traced back to the exact timestamps or transcripts, and reports ready to explore and share.

Ten years ago there was no version of the world where this would work. The big speed bottlenecks (like study creation, interviewing, and reporting) had ceilings. You couldn’t just go out and talk to all of your customers all at once. Now you can, in any language, at any hour, and in any medium, seemingly without compromising on quality, because today’s AI models are finally robust enough to be trained in the right way to ensure quality.

Listen’s traction has been encouraging. Customers range across startups and Fortune 100 companies like Microsoft, Alphabet, Robinhood, P&G, Nestle, NBCUniversal, and Anthropic. Using insights from Listen, Sweetgreen launched its popular Max Protein Bowl, Chubbies iterated on its kid’s swimwear line, and others ran studies on ads for the Super Bowl and Olympics.

In the nine months since they launched, Listen grew annualized revenue 15x, crossing eight figures. They have conducted over one million user interviews and raised $100 million from Sequoia, Ribbit Capital, Conviction, and others—including their $69M Series B announced in January. (Fun fact: Bryan Schreier, the Sequoia partner who led the deal on their side, was an early investor in Qualtrics.)

“Most of our growth now is inbound and word-of-mouth,” Alfred said. Listen’s AEs are back-to-back from 9AM to 9PM. “We’ve just started to build up our go-to-market machine.”

This is an impressive start. But how does Listen Labs win in the long run?

The strategy

Here’s how Alfred, the CEO, laid out the strategy:

“Phase 1: Displace legacy research vendors by providing a faster, better, cheaper way to provide customer research. Then democratize customer research by making it possible for companies to do it in places where they couldn’t before. Phase 2: Expand the platform to act on the insights in addition to collecting them. Phase 3: Become the de facto infrastructure for understanding people across all organizations, bringing forth a world that works in the way people want.

One piece that might help pull this off is a moat that most other AI companies do not have: Listen’s pre-qualified respondent pool of 30M people. This is the kind of thing that takes years and a lot of skill to build; you don’t vibe code your way to a pool of 30 million people in a weekend.

It’s partly this real, almost literal moat that sets Listen apart “from many other AI companies that are relatively thin AI wrappers,” Tobias Schindler, a founding engineer, told me. “We do the end-to-end user research process… That big AI labs would not get into.”

So Listen Labs may be somewhat insulated from one of the big fears around AI: that some upstart competitor could drink your milkshake simply by being clever, or that a company like Anthropic could trivially build their own in-house version and put you out of business. Listen has built a big, physical wall.

The early-mover advantage compounds in this space, too. “Every interview builds reputation scores on participants. We know who gives great answers. Every study makes the next one better,” Alfred said. There’s a clear flywheel for improving the product.

A final piece worth considering is the idea that Listen Labs’ success might create a Jevons paradox, where increasing efficiency increases demand. Companies don’t do as much research as they’d like to these days because it’s expensive, slow, and unreliable.

If research could become faster, cheaper, and more effective, though… The market could grow larger still. And that’d be a milkshake worth drinking.

The culture and team

Google had ping pong and foosball. Cursor has the whole “shoes off” thing. Walk into Listen Labs’ SF and NY offices and you will witness fika: a Swedish tradition where you’re meant to relax while enjoying coffee and pastries. Every Friday, the team gets together to welcome new team members, celebrate people who exhibited any of their six values (iterate to greatness, love the details, make it happen), and hear updates across sales, growth, and product. It’s also a time for Alfred and Florian to answer employee questions (they encourage “spicy” questions) that they solicit from the Listen platform.

Fika is the only big company meeting that Listen Labs holds. “Otherwise, meetings are almost never had,” someone told me. Engineers reported one to two meetings a week or less.

There’s another unusual internal communications norm: employees are encouraged to send DMs to one another in a channel called #private-dms. The thinking behind this is to create more transparency, better collaboration, and more accountability. Because these conversations are open and public, it lets others chime in and add context or support where they otherwise wouldn’t be able to in a private DM. This sounds a little insane until you try it, I was told, at which point you realize it helps people move faster and stay up to speed on what’s happening.

Oh, and they take their shoes off in the office, too. For good measure.

The obvious follow-up: “Who will I be enjoying my coffee and pastries with?”

Co-founders Alfred Wahlforss and Florian Juengermann met at Harvard. Florian is a German national champion in competitive programming and has worked at Tesla Autopilot, reporting to Elon; Alfred previously founded a YC-backed healthcare startup.

The broader team has the resumes you might expect (ex-Rippling, ex-Affirm, ex-Twitter, and so on). The team also counts three IOI medalists. When engineer Krish Mehta started on Labor Day, he expected to be alone setting up his laptop. Instead the whole engineering team was in, and Florian was remotely provisioning him access to every system from a trans-Atlantic flight. By day three, Krish was working on major infrastructure refactor and by day five, pitching new platform features.

Alfred and Florian know all too well that to keep accelerating their hockey stick growth, they needed to keep attracting top talent. As a startup, competing with AI labs and Meta on $100 million offer packages is a non-starter, so they’ve found creative ways to stand out. In March, they threw a Berghain-themed techno party to celebrate their Series B, hosted at a multi-story condo in SF, complete with a discerning bouncer. They invited customers, potential candidates, and friends of Listen employees. The goal was primarily to have fun together, but also recruit customers and top talent.

In 2025, they bought a billboard in SF and plastered a challenging engineering puzzle on it with no logo or text. The posts on X went viral, reaching over 3 million impressions. SF Standard explained the challenge: “The test asks participants to design an algorithmic bouncer for the legendary German techno nightclub Berghain…The player whose algorithm rejects the least number of people” won an all-expenses-paid trip to Berlin. 1,000 people attempted the puzzle, 60 solved it, and 1 engineer went to Berlin, on Listen’s dime. It was another way for Listen to build its employer brand and filter for top engineers in an unorthodox way.

Should you join Listen Labs?

Sometimes I find it’s easiest to decide whether you should join by listening to why other people at the company joined. So I asked Erik Bartlett, one of Listen Lab’s founding engineers (who was also one of the first few hundred people to work at Affirm).

“I was about to turn 30, and felt like I might as well take some big strategic swings by going very early somewhere; I could always go back to big tech in a few years if things didn’t work out. A longtime friend who was part of the talent team at Pear VC introduced me to some of his portfolio companies. Listen (then Merlin AI) was the first one I chatted with.”

Listen Labs is later-stage than when Erik joined—they are Series B—but still early enough for you to get meaningful equity and mold something life-changing if things work out.

You might also be a fit if you’re the sort of person who will do anything that helps the company, not just a narrowly defined list of responsibilities. “If you only want to work on your features and don’t debug customer issues, it’s probably not a great fit,” a founding engineer said.

Ria Shah, a growth lead, said Listen optimizes for “moving faster over doing things the ‘right way’. We’re a culture of being very quick to execute. The pace demands your best work.”

Listen Labs is currently hiring across most departments. If you want to get creative with your application, you absolutely can. “One of our recent hires shared ten really amazing ideas, which showed he spent a lot of time thinking about our space and was passionate about joining the team,” Alfred said.

You can read about some of our ideas for getting hired here, if you’d like to join what would be—if successful—a startup that changes how companies make their most important decisions.

Thanks to Listen Labs for supporting Next Play and making this Spotlight possible.

There’s a temptation to believe that just because something can exist, it does exist.

Humans went to the Moon (and are going again). We landed robots on Mars. We have quantum computers. We built the Large Hadron Collider. Inventing miraculous technology is our default.

Surely, then, we must have solved one of the most important things in the world—sleep?

But we haven’t. Or, at least before Eight Sleep, we hadn’t. Sleep is one of the highest-leverage pillars of health, and yet as Eight Sleep’s co-founder and CEO Matteo Franceschetti points out, most of us spend a third of our lives “on a dumb piece of foam.”

Turns out, there was a lot of space to innovate. In the 11 years since their 2015 Indiegogo Launch, Eight Sleep has built one of the most impactful sleep products in the world. They are now a unicorn, sell in 35 countries, and are the sleep system of choice for everyday people to professional athletes (like Charles Leclerc) and tech execs (like Mark Zuckerberg).

Their success so far is impressive. But the vision, Eight Sleep’s co-founder Max Bassi told me, is even bigger than sleep: “Eight Sleep will be the health company of the future.”

I was curious: what does that mean, exactly, and how is Eight Sleep going to get there? What is the product like today? What is the long-term vision? And who should consider joining?

The product

The way I’ve described Eight Sleep so far might imply that their core product is a mattress. But it isn’t. Eight Sleep’s main product, the Pod, is basically everything but the mattress. It’s a sleep system that works with your existing bed to improve your sleep.

Building the Pod meant solving problems no one had solved before: how do you track biometric data accurately without skin contact or a wearable? How do you adjust temperature independently on two sides of a bed, in real time, without waking anyone up? How do you do all of this without changing the comfort of a mattress? The result is a system of highly technical components that works so invisibly most people just call it “the thing that changed my sleep.”

The nice thing is that it actually works. Eight Sleep has conducted over 75 clinical studies to date with their internal research team. Their most recent peer-reviewed research shows that sleeping on the Pod clinically improves deep sleep, REM sleep, resting heart rate, and heart rate variability. And if any of your friends or coworkers are Eight Sleep customers you undoubtedly know about it; not because you asked but because they, unsolicited, will have told you all about how well they sleep. There are few products in the tech startup world that receive customer feedback with such conviction.

It helps to think of the product’s benefits in two buckets: (1) sleep and (2) health more broadly.

The former is the reason most people spend so much money on the Eight Sleep Pod. They want to sleep better. People want this because it is obviously good (sleeping is nice) and also because they know it will have other downstream benefits to their life and health.

The latter—health—is where things get even more interesting. What becomes possible when you sleep on a device that can track important health metrics, like HRV, respiratory rate, sleep stages, and more? What becomes possible when you have a clear picture of these metrics every single night for years on end?

These are the questions that Eight Sleep’s broader vision is aiming to answer.

The strategy

Eight Sleep is already cash flow positive. People love their product and pay plenty of money for it. If they wanted, they could stop at The Pod. The company could survive, even thrive, without another world-changing goal. Better sleep could be world-changing enough.

But the Eight Sleep vision is broader than a mattress upgrade.

“Eight Sleep will be the health company of the future,” Max Bassi, co-founder and CTO, told me. “Today we optimize your sleep. In 20 years, we cover you 24/7 - predicting changes in your body and health, and taking action automatically to help you achieve your goals, keep you healthy, and keep you well rested.”

The Pod, Max said, is the easiest entry point to getting the data you need to achieve this goal. People spend (hopefully!) 7 or 8 hours a day in bed, around the same time of day, every single day. They don’t move too much. Their body is relaxed. It is the perfect place and time to track all sorts of important health metrics; numbers that have wide-reaching consequences.

From this foundation, “we build outward - more sensors, more data, more interventions,” Max said. “Adjusting your temperature, changing your sleep schedule, alerting you before a problem becomes serious. The more nights of data we collect, the smarter the system gets, the more we can do for you. That’s the 20-year bet.”

You may have a separate question: how does Eight Sleep beat all of the other companies that want to play in this space? But the answer is more or less the same: the moat for Eight Sleep is the product they sell today. It is owning the bedroom, the best place to start with tracking many of the metrics that are important for preventative health.

The other differentiator may be Eight Sleep’s AI-powered products. One example is Autopilot, an algorithm that learns your sleeping patterns and proactively prevents sleep disruption. Another idea is AI-driven health guidance based on data that only Eight Sleep has.

And perhaps it is this grand vision that helps explain why Eight Sleep has raised $150M over the past year. Not only as a sleep system, but as a preventative health company.

The growth

The first version of Eight Sleep was a mattress pad on IndieGogo that they launched not long before being accepted into Y Combinator. The founders (Matteo, Alexandra, and Max) enabled the campaign at 2AM just to test that everything worked. A few minutes later, an order came in.

“We assume it was one of our test transactions,” Max said. “But then we looked at the shipping address. It was in France. A real person, on the other side of the world, had found us and bought our product.”

The company has done rather well for itself in the 11 years since. People like Mark Zuckerberg and Elon Musk have, organically and unprompted, talked about how they love the product. Athletes have purchased the product on their own, and then come to Eight Sleep wanting to promote it. “We don’t have to convince people to say they like it,” Max said. “They already do.”

Eight Sleep sells in 35 countries. They have hundreds of thousands of customers. They have collected over a billion hours of biometric sleep data. They are cash flow positive. All the kinds of things that most tech companies dream of.

As you might expect, VCs have been happy to fund Eight Sleep along the way. But the past 12 months have been the most dramatic flurry of fundraising in the company’s history. In August of 2025 the company raised a $100M Series D, led by HSG. And in March this year, they raised an additional $50M, this time at a $1.5B valuation, in a strategic round led by Tether Investments.

Why all of this fundraising, why now? I would bet that investor confidence is powered by a couple of things. One of them is financial fundamentals; Eight Sleep has built a winning product in the (notoriously difficult) consumer hardware space, which has done a lot to de-risk the business.

But likely the bigger reason that investors are excited right now is to help Eight Sleep accelerate towards their big goal: becoming the most important health company of the decade.

The team

What’s it like to work at Eight Sleep? It became clear through my conversations that Eight Sleep operates with a focus on discipline, effectiveness, and ownership.

Among others, I talked to Sam Kang, Eight Sleep’s Chief Product Officer who in the past worked on the Tesla Roadster and Sonos speakers. “There are three things I care about in a job,” he said. “The first is impact. The second is learning and solving interesting problems. The last thing is team… It was clear that the team was very passionate and smart, but I didn’t detect any ego.”

Meetings are scarce and the culture is “very written,” Sam said. “People add comments on agendas and then have discussions in meetings. Rarely are meetings just one-way presentations of information. One saying from our CEO captures it well: a good meeting is one that ends early, and a better meeting is one that does not take place.”

The culture is deeply focused on ownership of outcomes and accountability for your work. Vanity metrics, and the people who like them, don’t tend to do well. Nor do people who need a lot of structure or need to be told what to do. It’s hard work and the hours are long sometimes. Importantly, you need to know what your actual job is. In other words: what result am I being hired to produce? If you’d prefer to be handed a to-do list, Eight Sleep is not your company.

“The good is the company is full of extraordinary people, the mission is real and the product has the potential to improve the lives of millions of people,” Sam said. “On the flip side, expectations are high and the intensity is not for everyone.”

One other thing that sets Eight Sleep apart from other companies, David Q. Sun, VP of AI/ML (ex-Apple) told me, is “how relentless we are in asking, ‘What can be better?’ There is very little complacency, even when something is already working well.”

Should you join Eight Sleep?

There was one theme repeated by everyone I talked to: impact.

“You’re working on a physical product,” Max said. “Something you can actually touch, something you and your friends use every day. That’s rare in software. Most engineers spend their careers building things that live behind a screen. Here, your code changes how someone sleeps tonight.”

It’s true that in tech you do not get many opportunities to work on products that physically exist. It’s even rarer that one of those products is making an obvious, tangibly positive impact on people’s lives every day. Better sleep and better health are two clear positives.

You might also consider that now is a unique time to join Eight Sleep, because the company’s ceiling is not simply sleep optimization. Their goal is to become the most impactful health technology company of the decade, and The Pod is a useful foundation to help them do that. Joining now means building towards a vision with a higher ceiling than “sleep optimization” might at first imply.

If you do want to get hired, Eight Sleep’s CTO told me that he’d “be impressed by someone who sends me a project they built on the Eight Sleep API. It shows they understand the product, they can ship, and they care enough to do the work before anyone asked them to. That tells me more than any resume.” This advice likely applies across all departments at the company; show you can be valuable instead of telling.

The Eight Sleep team works both in-person at their San Francisco, New York, Boston and most recently, Milan, Italy offices as well as remote, depending on the position. See their open roles here.

Thanks to Eight Sleep for supporting Next Play and making this Spotlight possible.

It’s a bit crazy how fast the definition of “impressive” has shifted in software development. Twenty years ago it was impressive if somebody built a website, any website. Then Wordpress and website builders came along and a site was less impressive, but building an app was still a feat.

Now, someone could show you a full-fledged software application built for the enterprise and your reaction might be: “meh. seems vibe coded.”

It seems unlikely we have fully adapted to this new world yet. What are the consequences? What happens when the hottest companies in the world are using AI to write almost all of their code? When companies are shipping 2X, 5X, 100X faster? What problems arise?

One obvious side effect: testing becomes a lot more important.

There is more code than ever. Fewer people than ever are writing the code or even understand how the code itself works. This might be a bad thing. Today’s spotlighted company, Momentic, wrote about this problem in their recent Series A announcement:

“[Shipping faster] is incredible… Until you realize you’re shipping bugs faster too.”

Traditional testing (QA) tools were already brittle and flaky. They could barely keep up even when humans were writing all the code at a human pace. Nobody liked QA before, but it’s simply untenable now in a world where AI is doing a lot of the work.

So is building better testing infrastructure? If you ask Wei-Wei Wu, the answer is Momentic (the company he is CEO of). They raised their $15M Series A in November last year and want to become the “definitive verification layer for modern software.”

That sounds nice. But I wanted to dig deeper: what does Momentic’s product actually do, and in what ways is it really new? What is their trajectory? And should you consider joining?

The product

Historically there have been two ways to test software: one way is you write code (often in Playwright or Selenium) that runs automatically and tests things for you. The other way is you test the product manually, clicking through by hand. In practice most teams do some combination of both.

These approaches are brittle. They are the reason few engineers will tell you that QA is their favorite part of the job. A script that says “find the element with ID submit-btn and click it” breaks the moment a developer renames that element. And in the 22 years since Selenium came out—despite attempts at ‘automatic’ solutions—not much has changed. Testing scripts are still full of Xpath, data-testid, and hardcoded attributes that nobody cares to maintain.

The other problem with the old approaches is that they are slow. It takes time to write the tests, it takes time to run the tests, it takes time to maintain the tests. Far too much time for an era where code generation is automatic and a fast feedback loop is the most important thing to make sure you can actually ship a good product.

Momentic is taking a net new approach: you describe what you want and AI agents interpret your instructions against a website, iOS app, or Android app at runtime. No intermediate code, nothing to maintain.

“We don’t generate test code like Playwright or Appium,” Wei-Wei Wu, CEO, told me. “This is how we’re able to achieve incredible speed (<200ms per interaction) and world-class accuracy and reliability. Notion runs hundreds of thousands of tests a day with a 99.99% pass rate.”

The workflow is rather simple to imagine:

• You have an idea of a test you want to run.

• You write down that idea for Momentic.

• Momentic runs the test.

• You get your results and fix what’s broken (if anything).

That is the basic summary. The obvious follow up is: “Well, is it useful?”

Momentic is small, and early, but the results they’ve published seem good: Retool now releases 8X faster. Quora went from 7 hours of daily QA to 30 minutes. Mutiny reduced production incidents by 85% and sped up their release cycles. GPTZero ran tests much faster (and sped up their release cycles, too).

This is an encouraging start. But how do they become a $1B, or $10B, or $100B company?

The strategy

There are a few key bets that Momentic is making as part of their strategy to become the definitive layer for software verification. The outcomes of these bets will likely help determine whether or not this is a company you look back on in 5 years and think “I should have joined!”

The first bet is the one we touched on earlier: as AI writes more code and teams want to ship faster, testing is going to become a more valuable (and important) industry than ever before. More testing will be required, and increasingly that testing will be performed on software that no human had a hand in actually writing. A successful verification layer in this world is one that both moves faster and is more effective than existing testing infrastructure.

A second, related bet is Momentic’s novel approach to product. The biggest products in testing right now are code-first. Whether via AI or your own hand, you are writing code to test your product. Momentic is making a bet that the future of testing, of AI-native QA, will involve AI testing software at runtime instead of building brittle test code. The idea being that because this is faster and more effective, it will replace the status quo approaches to testing.

And then there is a third bet: that the QA function as it exists today will cease to exist. Instead, it will transform into something more integrated. “Roles across QA, engineering, and product are blending together,” Wei-Wei told me. “QA as a separate role and function from [engineering and product] is inefficient.”

With Momentic, people owning the product will be the ones owning testing. “Quality should be wholly owned by the person who built it,” Wei-Wei said. “It’s the fastest feedback loop and will only become faster and tighter as better AI models and tooling become available.”

To be clear, Momentic does not believe that QA people or their work are useless. Rather, they believe that today’s workflows are inefficient. That testing should be more effective and faster, that the feedback loops should be tighter. That QA should be better integrated.

The company is so opinionated about this shift that they don’t even sell to non-technical teams. “Our ICP is highly technical engineering teams,” Wei-Wei said. “A lot of startups in the space cater to the non-technical persona, which is a mistake. QA as a role won’t exist soon.”

There is logic to this. The current approach to QA at many companies is like if spellcheck didn’t live inside your word processor but you had to instead email it to Grammarly, and in a few days they would get back to you with the corrections. That would be inefficient. In QA, this kind of thing is the status quo.

The future Momentic is aiming for is one where:

• Testing is more important than ever

• Code-first tools aren’t good or fast enough

• Engineers own testing, not a separate QA function

In this world, companies will need a different sort of software—a new layer—to help them confidently ship as fast as they’d like to. If they win, Momentic will be that new layer.

The growth

In 2024, co-founders Wei-Wei Wu and Jeff An were accepted into Y Combinator’s winter batch of companies. Most of their first customers in this era—when things were small—came from launching often and talking about it on social media. A few were personal connections.

This rather humble start snowballed quickly. Customers started coming through inbound. Via Google Search. Via some paid ads. ChatGPT and Claude started recommending Momentic. Customers started referring to other customers. ARR began to grow.

In March last year, not long after graduating from Y Combinator, Momentic announced their $3.7M seed round. And just 8 months after that, they had raised a $15M Series A led by Standard Capital with participation from Dropbox Ventures (and existing investors like Y Combinator, FCVC, Transpose Platform, and Karman Ventures).

Momentic doesn’t share many other metrics, like ARR, publicly. (Wei-Wei told me that they are growing fast, which is perhaps unsurprising considering their two rounds in 2025.) But the customers they feature (Mutiny, GPTZero, Notion, Retool, and others) are impressive, as are the results those customers are purportedly seeing with Momentic.

There is a long way to go, but this feels like as good a start as any.

The team

Sometimes you read about startups whose founders have nothing to do with the problem they are solving. We’re two ambitious college grads who made a few SaaS prototypes and now want to create a brain scanning company! This can be fine. It can work.

But it is often more confidence-inspiring when the founders and the team do seem to have real experience in the thing they are solving. That is the case with Momentic.

Both Wei-Wei Wu (CEO) and Jeff An (CTO) are technical. Wei-Wei was the first employee and founding engineer at Nashi (acquired by Density in 2021) and has been an open-source contributor to Node.js, Storybook, Rollup, and more for years. Jeff built integration test infrastructure at Robinhood, then led enterprise quality at Retool (where he was recruited by Nicole Sgarlato, who is now Momentic’s recruiter).

Now that may sound good, but what would it be like to work at Momentic?

“Our edge is our culture,” Kailen Swain, founding product designer, said. “I’m sure every company says this so it sounds pretty stupid but here’s why: We’ve built a team where we can be completely honest, pointing out flaws and debating ideas because there’s a foundation of deep mutual respect. We win because we’re building with people we actually enjoy being around and who are all-in on the same goal.”

We often hear from startups about how the team trusts each other. But, at Momentic, trust is something that has to be earned. “Wei-Wei is not inherently trustful,” one team member said. “When I first started, he didn’t trust me, and I’ve noticed that he doesn’t trust anyone when they first start. He pushes back a ton in the early days. Once he trusts you, then he lets you run wild.”

I’m told that this early trust-building phase with leadership isn’t intimidating, though. Just the opposite—there is a lot of support as you prove yourself. “I know I’m hard to earn trust from early on,” Wei-Wei said. “So I try to be really explicit about what I’m looking for and give direct feedback fast, so people aren’t left guessing where they stand.”

Like you might expect with a startup of Momentic’s size, working there means you can expect to work on things that are not exactly in your job description. “I spend a lot more time at Momentic thinking about product and business goals [than I have at other jobs],” Henry Haefliger, a software engineer at Momentic, said. Still, though, this is an “engineering-first culture selling to engineers,” he said. “The people at the company are the right ones to build the product.”

The whole team works in-person in San Francisco. The culture is extremely outcome and output-oriented. One of the qualifying attributes for success listed on Momentic’s own career page is simply: “You dislike meetings.”

Should you join Momentic?

For now, Momentic is small. Smaller than most of the companies we spotlight here on Next Play (which tend to be at Series B, C, or beyond). Today there are 14 people on the team. Joining now means you’ll know everyone, probably quite well. You’ll be an important part of the founding team. You’ll help shape the culture, the direction, the roadmap. Being the third or fourth engineer (or marketing hire, or whoever) is different from being the thirtieth.

Perhaps most importantly, I think you need to both agree with and be excited by some of the ideas that (probably) need to be true for Momentic to succeed:

• There is more demand for testing in an AI era

• Current QA tools are frustrating and can’t meet this new demand

• QA as a function separate from engineering is inefficient

• Software is better when engineers own quality

• The best approach to testing is to verify it live

If these ideas ring true for you, the other question is whether Momentic is the right place to go build this. The early customer list—and those customers’ results—would indicate that Momentic is ahead of the pack on truly becoming the new layer for software verification.

“We’re growing fast,” Wei-Wei said, “and some of the best product teams on the planet, like Notion, Quora, Bilt, Webflow, and Retool, use Momentic every single day.”

If you are excited by the trajectory and buy into this vision, there would be no better time to apply to Momentic than right now. You can check out their open roles (all in person!) here.

Thanks to Momentic for supporting Next Play and making this Spotlight possible.

“Hello, IT. Have you tried turning it off and on again?”

This iconic line breached popular culture with the British sitcom IT Crowd in the mid-2000s. But talk to anyone who works in IT and you’ll quickly learn that the show did not invent the line. Rather, it was funny because it was true: IT people have been doing tedious work for decades.

• “How do I reset my password?”

• “Can I get access to [software]?”

• “The internet isn’t working.”

• “I don’t know how to log in to Slack.”

• “Is ChatGPT down?”

• “My new email account isn’t working.”

Requests like these make up a huge number of the tickets an IT team will get. If you need further evidence, read the top comment under the IT Crowd catchphrase video compilation: “After 3 years in IT, this line has resolved 50% of my calls. No joke.”

It’s not that IT people enjoy any of this. A 2017 survey ranked IT as the 10th most boring job on the planet. Nearly half of IT help desk agents quit their jobs every year. Nobody spends four years in college so they can spend the rest of their life resetting passwords.

And come on, it’s 2026! We have technology that can write a functional v1 of an app based on a single-sentence prompt! And yet we still have IT people resetting passwords? It still takes hours of human time to onboard a new hire?

Who’s trying to fix this?

The answer is Serval. Their pitch: the AI platform for employee support, starting with IT work. (But, as we will cover, the vision goes further than that). Serval signed Perplexity, its first customer, in May last year. They announced a $47M Series A in October, and raised a $75M Series B from Sequoia in December. Woah.

I wanted to know: what’s going on here? Why all this traction? And should you join?

The product

You may not spend all your free time researching IT Service Management (ITSM) tools. That might make it easy to underestimate the market. So let me enlighten you.

Have you ever heard of ServiceNow? It’s a software company founded in 2004. Sequoia was one of their early backers. The company grew and went public in 2012. Last year, ServiceNow made ~$13B in revenue. That’s more than Airbnb. More than Dropbox, more than Datadog, more than most of the startups you’ve ever heard about.

ServiceNow is the incumbent ITSM software platform. Yes, the market is huge.

Despite all of this revenue and incumbent success, however, IT workflows remain broken. The existing tools designed to build automations are more trouble than they’re worth. And few of them are AI-native. If you’re an IT person, why spend two weeks in a clunky drag-and-drop automation builder when you could just do the task manually in a few seconds?

Of course, this is not a good long-term plan.

Here is how Serval solves the problem: when someone in IT wants to automate something, they describe it in plain English. Like: reset Okta MFA factors, don’t allow this more than twice in 24 hours, require manager approval first. Serval writes real code from that description, which the IT person reviews and can publish. Then the task is automated for every employee, forever.

With Serval, IT people can write plain-language prompts to automate:

• Onboarding

• Offboarding

• Access requests

• Password resets

• Software license provisioning

• Any request that today is a Slack message to the IT team

A few real-life examples: Perplexity uses the platform to automate all of their onboarding and 50% of IT requests. Together AI automates 95% of just-in-time (meaning, time-sensitive) access requests. Mercor automates contractor onboarding. (These are just highlights; each of these customers automates ~100s of things with Serval.)

The product works either as a standalone ITSM or as a layer on top of existing software (like ServiceNow), a smart play since lots of legacy software is, for now, deeply embedded.

The strategy

Remember how I said the ITSM market is massive? How ServiceNow makes more money than most of the startups you have ever heard about? It’s true. But it’s not enough for Serval.

“We’re not limited to the ITSM or enterprise service management software space,” co-founder and CEO Jake Stauch said in an interview. “Yes, there are tens of billions spent on the software, but there are hundreds of billions spent on IT support and internal employee support… Our mission here is to build the AI platform for employee support.”

It’s not every startup that—even in its early stages—brushes past a tens-of-billions dollar market for something bigger. So Serval is aiming high, which is encouraging. But what’s the plan to get there? How do you actually become the AI platform for all employee support?

You start by solving one set of important problems, in Serval’s case IT. They have done this well and continue to do it well. Once you’re embedded within IT, you expand horizontally across the organization. Sequoia, which invested in ServiceNow decades ago and also led Serval’s Series B, said that “the last time we heard [customer feedback] that supports the thesis of [an] IT system of record empowering horizontal enterprise automation was 16 years ago, when we partnered with ServiceNow. That is why we were so eager to partner with Serval.”

There are clear signs that the AI automation market for IT is hot. Last year, two big acquisitions happened within a month of each other: ServiceNow acquired Moveworks and Automation Anywhere acquired Aisera. Both Moveworks and Aisera build AI agents for automation and could probably have been considered Serval’s two biggest competitors.

Even though their website copy remains IT-focused for now, Serval noted in their Series B announcement that their product is being “rapidly deployed beyond the IT department into HR, legal, and finance.” This is happening naturally; companies just recognize that Serval is useful and want to use it to automate anything they can.

Serval is still early. It would perhaps be too ambitious to make predictions about exactly where this strategy will land them, but one thing is clear to me: the early trajectory is strong.

The growth

The story surrounding Serval over the past few months is the kind of thing that most new startup founders fantasize about. In April of 2025, co-founders Jake Stauch and Alex McLeod still did not have a real customer. “Should we do something different? Should we make some kind of radical change to the business?” Jake remembers himself saying.

Weeks later, in May, they signed their first customer. And it was a big one: Perplexity.

Momentum built fast. June and July produced more customers and contracts, each bigger than the last. In October of 2025, Serval announced their $47M Series A led by Redpoint Ventures with participation from First Round, General Catalyst, and many others.

And then came the home run: on the day of the Series A announcement, a partner at Sequoia texted Jake one sentence: “come to our office.” Jake was in Orlando and couldn’t fly out, so Sequoia came to him, and it wasn’t long before they had a term sheet. In December, Serval announced a $75M Series B led by Sequoia at a $1B valuation.

Less than a year ago, Jake and Alex were in the same position many startup founders find themselves in: is this even going to work? Now their company is a unicorn.

If you are a skeptical reader (perhaps for good reason!) you may start to doubt this story. You may wonder if this is all hype; if investors just have FOMO. But it seems like this investor excitement for Serval is based on actual facts:

• Serval grew revenue 500% between August and their Series B.

• They automate ~50%+ of IT tickets across their current customers.

• Customers are—without being prompted—deploying Serval across the whole org.

It’s true that not all the numbers are there. We don’t know their ARR. We don’t know about any sort of plan for profitability. We don’t know their churn, and so on. Things could certainly still go wrong. But the start is promising and, with $127M total raised at a $1B valuation, it seems like Serval has plenty of fuel to see just how far they can go.

The team

Serval’s story—and some of their founding team—starts at Verkada.

Co-founders Alex and Jake, along with founding engineer Sebastien deGroot, worked together at Verkada doing all sorts of automation (among other things). “When Jake and Alex left,” Sebastien said, “the cultural shift was obvious. We stayed in touch, and about six months later I visited them in a small WeWork to see what they were building. It already felt like a novel product with real potential.” Serval’s COO, Tatiana Birgisson, did not work at Verkada but previously led growth at Rippling. It’s certainly a confidence-inspiring lineup.

When I talked with the rest of the team (who seem equally impressive) about work at Serval, it quickly became obvious that this is an opinionated company where you can expect to work really hard, even by startup standards.

For one thing, there are no product managers, Sebastien told me, with Serval opting for the Palantir-esque model of hiring forward-deployed engineers that build product through direct customer feedback. This is part of why, as engineer Derek Zheng said, “customers can expect turn-arounds on feature feedback within the same week and, often, the same day.”

Of course, working at a place where customers can expect turnarounds on feedback within the same day is an intense experience. Spencer Lang, an FDE at Serval, told me that his first week he was “immediately thrown into customer calls and shipped core features like emails from our workflow builder and our Crowdstrike integration.” The second week of work at Serval, he said, the company became a unicorn.

Practically-speaking, you can expect to work long days and—according to a few people I talked to—sometimes weekends. “It is NOT a traditional 9-5,” Helen Taylor, a mid-market AE, said. “Anyone who needs structure and concrete processes may not enjoy working here.”

Everyone I talked to said they worked for at least 10 hours a day, and the most dramatic answer was pretty succinct: “I work 15 hours a day because I love it.”

This isn’t to say that you will work 15 hours a day and weekends if you join Serval. But I doubt it’s the right place to go if you are in a four-hour work week era of your life. You will work hard here!

Luckily, you will work hard alongside some truly talented people.

Should you join Serval?

“During my last interview,” one engineer told me, “I asked one of our founding engineers why [he joined] Serval. He told me that if he thought he were better than the founders, he would start his own company. This really stuck with me because I thought he was one of the strongest engineers I’d ever met.”

There’s a lot of love at Serval for Jake and Alex and the team they’ve hired so far, and so one reason you might join is because you feel like the engineer in Spencer’s quote: I trust these people to run a great company more than I’d trust myself.

You might also join because enterprise automation (despite sounding like a buzzword) is actually quite important. On the human level you are saving IT people and legal people and HR people from doing tedious manual work. And on the business level you are helping companies pay people to do more valuable, higher-leverage work, which helps them grow faster.

If you are convinced (or semi-convinced) that Serval is going to achieve even a fraction of its grand vision, then now is the time to join. There is no better time to get equity in Serval than now—a phrase that will be true for as long as the company continues doing well—and they are early enough that an exit or liquidity could result in life-changing money for you. (This is often less true for joining a startup at ~Series D and beyond unless something crazy happens.)

Finally, I think you might join so that you can help realize a new version of that quote:

“Hello, IT. Don’t worry. Serval will take care of it.”

You can browse Serval’s open roles here.

Thanks to Serval for supporting Next Play and making this Spotlight possible.

Here is a T.V. show pitch for you: we reboot Sherlock. Bring Benedict Cumberbatch and Martin Freeman back. Rebuild the old sets, hire the old writers. It’s detective time.

There is one catch, however. The first eight episodes are just Sherlock Holmes sitting behind a desk doing paperwork. That’s it. Each episode is one single shot; the camera doesn’t move.

Then, in the last couple of episodes, Sherlock does some real deducing and solves the case.

What do you think? Riveting, no? Maybe I’ll send a cold email to HBO.

Of course, this T.V. show would be bad. You wouldn’t watch it. If I got HBO to make this show, it would be ridiculed as the biggest waste of money since the final season of Game of Thrones. Sherlock Holmes is a detective, and he’s supposed to be doing smart detective things—not tedious manual paperwork that anyone could do. I am sure you agree.

Now what if I told you there is a real version of this T.V. show happening every day?

It’s not Sherlock Holmes. It’s financial crime investigations. Every day, thousands of smart analysts working at financial institutions from Robinhood to J.P. Morgan wake up, open their computers, and spend the majority of their time doing repetitive pattern-matching. Only a small fraction of their day is used for actual deduction about whether crime has happened or not.

Financial crime is a huge deal and it’s seriously underrated relative to how much of a role it plays at financial institutions. In a 2024 letter to the FDIC, the popular fintech startup Mercury said that 1 in 4 people at Mercury work in compliance (much of which is about financial crime). And that’s not out of the ordinary; North American companies spend $60B+ per year on financial crime compliance.

The world may be a much better place if the following things happened:

1. Financial crime investigations become far more effective (less fraud)

2. Financial crime investigations become far more efficient (humans focus on better work)

3. Financial crime investigations become far more flexible (labor cost scales to workload)

If you pitched this thesis to Will Lawrence, founder and CEO of Bretton AI, he might tell you that the solution is simple: have every financial institution use Bretton AI as their infrastructure layer for financial crime investigations.

Last month, Bretton AI (formerly Greenlite AI) raised a $75M Series B to help financial institutions automate financial crime investigations with AI agents. Below I’ll share more about how it works, what the long-term play is, and what kinds of people should consider joining.

The product

Startup founders often enjoy describing their companies as the “AWS of [X].” I have heard about AWS for the internet, AWS for healthcare, AWS for AI. Perhaps we are not too far off from a future where some founder pitches the AWS for AWS. It wouldn’t surprise me.

And so when founder and CEO Will Lawrence told me that Bretton AI is like “AWS, but for compliance workforces,” I raised my eyebrows. But as I dug further into what Bretton AI does and what the world looks like if they succeed, the comparison felt apt.

The product’s outcome is simple: it automates the bulk of financial crime investigation work for financial institutions. Existing customers include fintechs like Mercury and Robinhood to larger, FDIC-regulated banks; really any company that has to deal with financial crime compliance.

So how does it actually work?

When Bretton AI graduated Y Combinator in the summer of 2023, it had a narrower scope: it was an AI tool to automate sanctions screening reviews. Sanctions screening is what you have to do when a customer’s name matches against a government watchlist. You, the person doing the review, have to check and see if the alert is correct or not.

Will Lawrence and his team knew they wanted to tackle all of financial crime and compliance, but sanctions screening was the place to start. It is fairly straightforward (relative to some other kinds of compliance work), but it is also important and requires a lot of manual labor. And the competency of AI models in 2023 was enough to automate this effectively.

Fast forward to 2026 and Bretton AI has hit an inflection point. “First,” Will told me, “LLMs crossed the threshold where they can reliably handle complex, judgement-intensive compliance work. Second, regulators are increasing pressure… So compliance teams are drowning. Third, the talent market for compliance analysts is brutally tight and expensive. Banks are spending billions on compliance staffing and still falling short.”

This is the perfect storm for Bretton AI’s current product, which operates as a more comprehensive suite of investigation-ready AI agents that can tackle almost all portions of financial crime compliance. Two of the biggest things it does today are AML (anti-money laundering) and conducting enhanced due diligence (like KYC and KYB), both of which today require tons of tedious, repetitive manual work.

With Bretton AI, a company’s compliance analysts can focus on the Sherlock Holmes-y part of the job, not the database searching/copy-and-pasting/repetitive pattern-matching part.

The AWS comparison slots in nicely via the mechanics of how companies use and pay for Bretton AI. Instead of flexing analysts up and down with demand (complicated), companies can scale capacity up and down immediately (easier). Like how you’d scale compute.

And, to be clear, the product works. In their fundraising announcement last month, Bretton AI shared that:

• A $15B+ institution cut BPO spend by $5.35M in Year 1.

• A Fortune 500 reduced time to onboard institutional clients by 50%.

• An FDIC-regulated bank reduced time to originate loans by 90%.

This is a promising trajectory. So where is it all headed?

The strategy

If Bretton AI works out, then in twenty years “it becomes the infrastructure layer for staffing in regulated industries,” Will Lawrence said. Bretton AI is capitalizing on a theory that many startups (and analysts, and journalists, and others) espouse: regulated industries are filled with a ton of tedious work humans shouldn’t have to do.

Nobody wants to watch the ‘Sherlock Holmes does repetitive paperwork’ show the same way nobody wants to be the Sherlock Holmes doing repetitive paperwork. Financial crime is interesting and important, but people want to spend their time doing the tasks that really require a human; the complex tasks and the deduction. The Sherlock stuff.

But Bretton AI is starting with arguably the most difficult regulated industry first. Financial crime compliance is so important to financial institutions—and the consequences so large—that convincing an FDIC-regulated bank to use your product is different from selling to a small startup.

That’s why Bretton AI started with fintechs. Mercury, Robinhood, Gusto, and many others; these startups were willing to take a chance, and luckily, the tools worked. Now the question is: how does Bretton AI move upmarket?

Two things stand out. First is what Bretton AI calls Trust Infrastructure, a governance layer that embeds specific federal banking guidelines (like OCC and NYDFS requirements) into every agent, with full audit trails and explainable outputs. This is great because banking execs don’t need convincing that AI could help, what they want is proof it won’t screw things up in front of a regulator. Bretton AI’s Trust Infrastructure goes a long way towards doing that.

Second, Bretton AI is pushing into more complex investigations, and they are doing so faster than their competitors. Just last week they announced Horizon, a product designed to tackle what’s called L2 work: longer, judgement-intensive investigations that today take hours. “Our competitors have not cracked this,” one employee told me. “We want AI to not only automate low-risk work but also to build the best UI possible for [the complex stuff].” If they pull it off, that’s a meaningful moat.

The growth

It took 2 years between Bretton AI’s acceptance into Y Combinator and their $15M Series A. And it took just 9 months between that Series A and last month’s $75M Series B, led by Sapphire Ventures with participation from Greylock, Y Combinator, and others.

It’s not hard to see why Bretton AI is having success with VCs. Their average contract size has increased to $201k, up from $85k in May last year and $25k at their seed round. The total market cap of customers is now over $1T. Customers are no longer other startups but also include OCC-, FDIC-, and Federal Reserve-regulated banks. Bretton AI’s AI agents have now completed more than 1.2 million L1 and L2 financial crime investigations.

There are some people who say that AI is a bubble. These people may point to the less-exciting stats about AI products, like the MIT report last year that claimed 95% of AI pilots were failing. Perhaps some of these claims are true; perhaps there are a lot of AI tools that promise a lot but can’t deliver. Or tools that end up being a distraction, costing more than the value they deliver.

All the hype (and the backlash to the hype) makes it difficult to see through the mist and figure out which AI companies are actually helping their customers make more revenue. But Bretton AI’s numbers are about as positive as any that their product is delivering actual value.

And if Bretton AI ends up realizing its vision to become the infrastructure layer for all compliance in regulated industries—then the company has plenty more room to grow.

The team

I sat down with the team (which includes ex-Uber, Palantir, SpaceX, Stripe, and Airbnb employees) to learn more about their culture and who might fit in. What’s the vibe at Bretton AI? How does work get done?

Well, if you apply, you can probably expect to work in-person (though there are some limited remote roles). “You don’t realize how much gets lost in async until you’re back in person full time,” Nitya Mummalaneni, AI ops specialist at Bretton AI, said. “Being together every day means conversations don’t get lost in Slack threads, decisions are made faster, and connections feel more authentic. It’s life-changing for sure.”

Bingqing Wu, a product engineering lead, said she was excited to join because things are moving quickly. “I was especially struck by how much people are encouraged to think creatively and tackle challenges from different angles. One thing that might be ‘unpopular’ in the broader tech community is how much we value being practical over hype. The culture here leans toward substance over flash. We focus on solving real customer problems.”

People who work at Bretton AI are expected to “own outcomes,” and working at Bretton AI is not a “chill job,” James Rogers, a product manager at the company, said. But, for the right person, that’s a good thing. James started working at Bretton AI while wrapping up university and “ended up just skipping almost all my classes to work on the product.”

I also asked: who isn’t a fit? What would a smart person who quit Bretton AI say? Someone told me that you might quit because “there’s a lot of ambiguity and not a ton of process.” This is true of a lot of startups (especially earlier-stage) but it is worth reinforcing. Bretton AI is not a calm place to walk in and expect a well-structured, pre-defined list of everything you will be doing this week or this quarter. You’re going to have to try harder than that.

Should you join Bretton AI?

Perhaps Bretton AI isn’t too unrelated from Sherlock Holmes after all: 17 of the 56 original stories by Arthur Conan Doyle feature some kind of financial crime. After all, if you want to run a successful criminal operation at any scale, you probably need to do some illegal money stuff.

So, if you’re looking to be pitched on the mission: Bretton AI helps reduce real-world crime.

If you zoom all the way in, working at Bretton AI is in part about making financial investigations faster and more effective. You may play an important role in reducing the amount of financial crime in the world. You may also help to make compliance work more fair, a useful goal in an industry where discrimination and bias is still somewhat baked-in.

If you zoom all the way out, then working at Bretton AI is—as CEO Will Lawrence told me—about building an ‘AWS for compliance.’ About building a brand new infrastructure layer in one of the most important industries in the world: finance. If the company ends up becoming an embedded infrastructure layer in regulated industries, then joining Bretton AI now means joining a startup that leaves a real footprint on the world for decades to come.

The money could be nice, too. Bretton AI is at Series B, a good stage for a startup to be at if you’re interested in long-term equity (the risk-reward tradeoff is often favorable). They are on a fast upward trajectory. And if the company succeeds with everything it would like to, you would probably walk away with a lot of money. I’m sure you wouldn’t be upset about that.

If you would like to join Bretton AI at their San Francisco office or look for a remote role, you can do so here. And feel free to send Will (or someone else) a cold email. “If you cold email me with a POV on how we should think about a compliance workflow, or you’ve built something that shows you understand the problem, that goes straight to the top of the pile,” Will said.

(You can read some creative ideas for getting hired in this essay.)

Thanks to Bretton AI for supporting Next Play and making this Spotlight possible.

All industries are disrupted, but some industries are more disrupted than others.

You could even say that some industries are too disrupted. Product management. Productivity. Workplace communication. Documentation. Great new companies will probably crop up in these spaces in the future, but they aren’t where I would send the world’s most promising startup founders to go build.

Then there are other industries—unsexy but enormous—that have seen comparatively little disruption. We’ve covered industries like this before: like with our spotlight on Hearth, a company solving revenue for home service contractors, or Fulcrum, software for insurance brokers.

Today, we are covering a company that deliberately chose the hardest version of this thesis. Valon started in mortgage servicing, a $25B+ market encumbered by regulatory complexity, where the casual innovator is quickly deterred by the barriers. Six years of learning the regulatory complexity firsthand has compounded into real scale.

Today, Valon has:

• Built a top 10 mortgage servicer in the US, with over $110B of mortgages serviced on their platform

• 3 of the 10 biggest servicers signed up to deploy ValonOS

• Raised more than $230M from a16z, WestCap, and others

• Team of 200+ talented people

If you have any interest in learning about (and maybe even joining) quietly dominant startups disrupting unsexy but important industries, well, this spotlight is for you. I sat down with the team at Valon to understand how they built such an impressive business and where it is headed. (As a hint: Valon’s mission is to build the AI-native operating system for all of institutional finance, not just mortgages.)

I also asked the questions that you may be especially interested in: What is it like to work at Valon? And who, specifically, should consider applying?

The most common loan in America

You probably don’t need much convincing that mortgages are a big deal. Most Americans own a home, and most homeowners have a mortgage; more than $13T of consumer debt is underpinned by mortgages. And the way that mortgages are distributed and run can have major impacts on both regular people and the broader economy as a whole (see: 2008).

So, in some sense, you may know a lot about mortgages. But there are a couple of things you may not know that you will need to understand if you want to get a grasp on Valon’s business.

The first thing is that lenders are often not the ones who manage or ‘service’ a mortgage. Lenders often hand off the work of actually running a mortgage (payments, compliance, communication, etc.) to a kind of company called a “mortgage servicer.” Valon built their own (creatively named “Valon Mortgage”).

The second thing is that servicing mortgages has become significantly more complicated in recent years. The aftermath of 2008 brought a wave of regulatory crackdowns, which is part of why the cost of servicing a mortgage since then has tripled; we’re talking about a $25B market.

How did those loans get serviced before Valon? “With technology that was invented before the internet was created,” co-founder and COO Linda Du said in an interview. The clunky servicing software and high servicing costs were, naturally, problems being passed straight onto everyday Americans.

This reality, and the mission to fix it, deeply resonates with the people I talked to on the Valon team. “Mortgage servicing is what banking was before banks had mobile apps or mobile check deposits,”. “We’re rebuilding the entire core infrastructure in a way that is aligned with where the world is going: AI-native, automated, and with safety nets.”

If you were living in the pre-Valon world and wanted to build a startup, the facts would start to add up:

• Real problem that affects real people in painful ways

• Untouched industry with stone-age software

• $25B market that powers homeownership in America and continues to grow

What’s not to like?

From friends-and-family loans to $110B in mortgages

Back in 2019, co-founders Andrew Wang, Jon Hsu, and Linda Du saw a lot to like in the mortgage market. The origin story is classic startup lore: Andrew, who had been working on mortgage problems from the other side of the table at Soros, kept running into how messy and inelegant the whole system was. He wanted a better solution to exist, badly enough that he sketched out the idea on the back of a napkin with Brian McGrath, who was an early investor with Jefferies.

And they started with the same question many startup founders find themselves asking: Why can’t we just build a better piece of software to solve this?

What they discovered next was both humbling and, in retrospect, clarifying. A couple of impressive résumés and fancy degrees don’t conjure a perfect piece of software into existence — and the existing mortgage servicing companies, slow-moving by nature, weren’t exactly lining up to pilot something new. But there was a more interesting revelation buried in the difficulty: nobody else had tried to solve this problem not because it was a bad idea, but because it was so extraordinarily hard. The opportunity was sitting right there in plain sight, and the whole industry had collectively shrugged and walked away. That complexity wasn’t a warning sign — it was the moat.

So, instead of wasting time building unproven software nobody would want to adopt, Valon’s founding team asked a different question: What if we just build a mortgage servicing company instead? Most startups pursue a wedge strategy, starting small and expanding outward. Valon intentionally did the reverse.

Things started scrappy. In order to get a license to operate as a mortgage servicer, Andrew said in an interview, you’re required to have a track record with credit agencies. This was something of a paradox, so the team “found which states had exemptions for small loans, then went out and gave loans to a bunch of friends in order to be able to report 100 loans.” All of this just to get licensed.

The business snowballed, adding billions of dollars in loans. And as they grew, Valon built out the software they had envisioned from the start; effective internal software for servicing loans as well as great consumer-facing software for managing those mortgages.

The team calls this the “anti-wedge” approach and it’s the only way to deploy AI into regulated industries. Owning the operating system means every workflow and data trail is structured, inspectable, and programmable, making it the perfect surface for AI.

Fortune, or perhaps just a lot of hard work, has favored Valon. Valon’s mortgage operating company is now one of the top 10 mortgage servicers in the United States and has serviced over $110B in mortgages via its platform.

In some ways, things have come full-circle. Now that Valon has built a proven servicing software solution, they are also able to work on realizing the idea they started with: licensing their software to other companies. The broader vision is even bigger: “Valon is building the AI-native operating system for regulated finance—starting with mortgage servicing.” While the business is currently focused on mortgage-related products and has not publicly launched anything in other areas of regulated finance, this is worth noting; Valon’s ceiling is not the size of the mortgage market. It is much higher than that.

If you were searching for somewhere to work, all of this information is (probably) music to your ears. There is still a pending item, though: What is it actually like to work at Valon?

What it’s actually like to work at Valon

Ask most people at a startup what the culture is like, and they will give you aspirational truisms (we work hard, we move fast). These are not helpful if you want to know what it is actually like to work somewhere.

So what is it like at Valon?

Valon has offices in New York and San Francisco (though there is flexibility for some remote roles). As for the vibe? Understanding the culture is easiest by learning how people actually work. “Even as an intern I worked on meaningful projects with real ownership. The internship project I led generated roughly $1M ARR, and I was treated like any other engineer,” one employee said.“Leadership, engineers, product, biz ops, and analytics are all visible and approachable. Oh, and people bring treats from their travels.” One product manager brings loaves of bread on Friday because the team has to ‘get this bread.’

If you worked at Valon, you would find the kind of fun, intense energy that you may have had “back in college working with friends on a really hard problem late at night, bouncing ideas off of each other, furiously marking up a whiteboard.”

“Your coworkers are deeply bought in. There’s no coasting, no political nonsense. People take real ownership and hold a high bar for themselves and each other.”

Every Tuesday, you could expect laptops-down “family-style dinners.” “The week of Thanksgiving, we did a mini Valon Thanksgiving dinner that was a ton of fun.” But the culture of access extends beyond the dinner table. Andrew holds regular office hours — with a particular focus on AI adoption — opening his calendar to anyone in the company who wants in. It’s a small thing that signals something bigger: even as Valon has scaled, it operates with the accessibility of a much smaller startup. Anyone can get time with the CEO. And when the CEO is literally in the weeds with his team on AI, the whole company moves faster because of it.

This was refreshing to hear from the Valon team—it’s not too often a startup grows like Valon has and is able to retain this kind of atmosphere. “If you think you’re moving fast, know that you can always find ways to move faster.” “Fast > perfect.” Like at most (good) startups, at Valon you are expected to perform at a high level and to put your all into the work. This isn’t some sort of 996 model where everyone’s in the office on Saturdays, but you will be asked to meet high standards.

Valon is also not the kind of startup where teams work on pre-established playbooks; where anything you might do is defined six months in advance. Not long ago, some folks on their recruiting team built custom Monte Carlo simulations for pipeline forecasting and a Chrome extension for resume screening. Not because anyone told them to, but because they saw a problem and solved it. If you work at Valon, you can expect to be proactive and build things because they are useful and generate value—regardless of whether someone told you to do it.

“We aggressively kill meetings that are not useful.” Meetings at Valon are usually designed either for alignment or debate. If a meeting accomplishes neither, it gets cut.

One other theme I heard repeatedly: Valon feels like an asymmetric bet. The downside is genuinely capped (profitable, real product market fit, and a very strong balance sheet), but the upside is massive. Valon has resisted the hype cycle of chasing the latest AI fad and the inflated valuations that come with it. Employees joining now can still capture meaningful upside at multiples that are impossible at companies where growth is already priced in..

Two of the co-founders (Andrew and Linda) come from investing backgrounds, which is partly why they got to see what happens when founders and CEOs are chasing the latest hype cycle instead of thinking long-term. “Andrew and Linda are both very focused on creating long-term value. From my experience, they are better execs than most of the public CEOs I met when trading stocks at Millennium (a top tier hedge fund).”

“I was coming from a company with an unproven business model (and frankly no PMF). It was so refreshing to talk to a founder and company with a clear identification of the 3, 5, and 10 year strategies and numerous back-up plans to make money and win the market.

“It was nice to know I had to focus less on the strategic direction of the company and could pour my focus into the pieces I was personally responsible for.”

Should you consider joining Valon?

Valon is currently hiring for dozens of roles here. (There are two branches of the company—Valon Mortgage and Valon Tech—but if you are reading this you should probably look at Valon Tech.)

I’d start by asking yourself whether you’re the type of person who gets excited by untangling complexity that scares everyone else off.

“The competition is outdated, slow, and stuck on airport-terminal GUIs and we’re building something dramatically better in a massive, unsexy-but-enormous market,” one person said.

Valon told me that people who do well at the company tend to share three qualities:

• Turbo: relentlessly outcome-oriented, with the genuine enthusiasm to go 10 layers deep on a problem

• Anti-hype: motivated by creating actual value rather than chasing shiny things; comfortable playing the long game

• Precise: holding an extremely high bar for technical accuracy in systems where trust is a one-way door

So, should you consider joining Valon?

It may be easiest to answer that question by defining who shouldn’t apply. Don’t apply if you’re joining for a quick payout or a flashy headline; the payoff here will be massive, but the tradeoff is that it won’t be quick. Don’t apply if you need prebuilt systems and clear playbooks. No one here is a cog executing the machine. The job is to find the most valuable thing to do, and then go do it.. Don’t apply if your solutions tend to break under pressure; at Valon, systems have to handle zero downtime, strict compliance and edge cases that would collapse most software.

From a numbers standpoint, Valon is in a sweet spot: growth is inflecting but the company is also mature enough that you have a good chance at liquidity with the equity you get, but not so late that your equity may underperform the market. “The downside is genuinely capped: we’re profitable, we have real product-market fit, a strong balance sheet, and customer demand that’s already contracted or imminently landing. This isn’t a science project or a hope-and-pray startup. But the upside is massive,” Ariel Brito said.

“Valon offers an exceptional opportunity to grow quickly in your career, the company is at a pivotal stage, preparing to scale rapidly across many customers. If you want true hands-on experience in scaling a business, this is the ideal place to learn.”

If you do apply, what should you expect?

“The interview process was rigorous but fair,” a software engineer told me. “I did technical interviews focused on practical coding and system design, and I also had conversations with leadership about the company missions and goals. I really appreciated that people took time to answer my questions and make sure it was the right fit on both sides.”

You can find Valon Tech jobs here. Don’t be scared of reaching out directly if you see somewhere you think you can add a lot of value. If you do, consider reading our guide to getting a job with cold email before you do so to increase the odds your email is a good one.

Thank you to Valon for supporting Next Play and making this spotlight possible.

If you are a person who has been in regular meetings for the past few years, you have probably heard of Granola. And if you have heard of Granola, you may know the general consensus about it: Granola is the best AI notetaking tool for meetings. It transcribes your meetings, writes helpful notes, enhances your own notes… The sort of stuff you would’ve needed a real assistant for.

There is one version of the world where Granola is truly just that: an AI notetaker. If co-founders Chris Pedregal and Sam Stephenson wanted to, they could probably thin down the team to ten people or so, never raise another round, and run Granola as a profitable lifestyle business that owns its niche.

So it may surprise you when I say that Granola’s vision (and the reason they raised a $43M Series B from Nat Friedman, NFDG, and others last year) is far bigger than AI notes. Co-founder Sam told me:

“We want Granola to be the primary interface where knowledge workers do most of their work.”

If you are looking for interesting startups to join, this sentence should probably make Granola significantly more interesting. Now we’re talking about a many-billion dollar market. Now we’re talking about a trajectory that would position Granola as one of the most impactful startups of this decade.

Then again, you probably just know Granola as a notetaker.

So… how? What is the plan here? And should you join to help Granola accomplish this mission?

Don’t underestimate the scribe

Every software engineer knows that modern code editors (IDEs) are spectacular. They can help engineers work far faster and make fewer mistakes. And the AI versions these days could be the best yet.

On their pre-launch 2024 landing page, Granola wondered:

Why is there no great IDE—no great workspace—for knowledge work?

Today’s knowledge work is “fragmented across dozens of different tools,” co-founder Sam Stephenson said. “You’re constantly switching between email, Slack, your CRM, task management, [and so on]. We believe AI agents will eventually handle much of that context-switching and updating for you.”

You may have seen some early attempts at this in the market. Email software like Gmail and Superhuman these days can try (and often fail) to write good emails for you, as one example. Anthropic launched Claude Cowork back in January. Companies are aware that AI could integrate into everyday work.

Granola has had this in mind ever since they began working on their product back in 2023. The key question, though, was how to realize this vision in a way that would be actually useful. How do we make AI feel like a valuable extension of yourself and not a context-less slop machine?

Sam says the answer was simple: “For [this] to work, you need something that understands your context deeply. And meetings are where the most important context lives.”

If you were a medieval ruler with a royal court, you would want your advisors and scribes to be present for everything. You would want them to have the same information you do so that they could help you make good decisions. And you would trust them over a know-it-all who rode in from afar with no information.

Starting as a humble meeting scribe, then, was the smart place for Granola to begin.

The Cambrian explosion

Publicly, all was quiet for the first year that Chris, Sam, and their small team worked on Granola.

Internally, the product was undergoing an iteration process that Jim Fisher (software engineer) describes as a Cambrian explosion approach. The traditional v1, v2, v3 path was ditched for an iterative process in which they would onboard users one-by-one and ship all kinds of different permutations of the product; different features, different styling, different everything. As 2024 approached, they honed in.

And then, when Granola launched in the spring of 2024, they removed 50% of their features. “If you add 50 buttons in there with new features,” Chris Pedregal said, “you kind of kill the golden goose.”

This focus on restraint, taste, and what I would call “actual usefulness” paid off. Word-of-mouth took off and Granola was quickly recognized as the most reliable AI notetaker around. And people stuck around.

“From very early on, we had this retention curve that just flattened out at a really high level. People who did one meeting with Granola kept coming back. We didn’t have to growth hack our way to virality - once people tried it once, they just naturally kept using it for all of their meetings,” Sam said. 70% of people stuck around for a week. 50% of people stuck around for ten.

You know it’s good because an AI notetaker isn’t exactly an impossible product to build. (This isn’t powerful military tech.) I bet you could go spend a few hours with Claude or ChatGPT and vibe code a meeting notetaker this afternoon, and maybe it would even work. As a result there are 100s of legitimate competitors for Granola; the idea feels simple and easy to replicate.

And, yet, Granola is winning. They are winning a lot. This says something about the team’s approach: the restraint, the taste, the knowing how to build and what to build and when.

There’s a long way to go still, and “[the founders] don’t feel like we’re there yet: they want to keep listening to users and keep trying to think about how to make the product better,” Vicky Firth said.

Sam and Chris are “down-to-earth, optimistic, and visionary,” someone on the team said. “They are super transparent - a lot is shared about how they’re feeling, what investors are saying, how the company is doing. They have strong opinions about the product but place high trust in the team.”

Traction? Check. Smart, measured founders? Check. What about the culture?

Merge first, ask questions later

Granola’s offices in London and San Francisco, where everyone works in-person, are not safe spaces for complainers. “There is a huge amount of positivity and very little complaining at Granola,” Vicky told me.

“This covers everything from big product changes all the way to which drinks are in the fridge. There’s a buzz here that you don’t see everywhere.”

There is a “just give it a go” energy at Granola. You’re trusted to do a great job and, if you try something but it doesn’t work (like when Jim, an engineer, deleted half of the prod database), the team is not going to bury you for it but rather solve the problem and move on (and maybe ask you to drink ‘OMG whisky’.)

“There are no pre-code merge reviews,” an engineer said. “You just merge and iterate! Many other companies I’ve worked at would balk at this.” But it works at Granola, and this general philosophy—just do a good job and we’ll iterate as we go—seems to be present on every team.

How long should you expect to work? Everyone I talked to said they worked about 9 or 10 hours per day, but they also stressed that hours can be flexible. “You’ll often find me dashing off at 5pm for another commitment,” someone said. My overall impression is that, while the team at Granola works hard, this is not a startup that will ask you to come in on weekends; they will not force you to stay up until 2AM working on some feature (though I’m sure you could do it if you wanted to).

Shreman Shrestha, who works on various business and growth projects, had three points to sum it up:

“1. Highest density of talent I’ve ever encountered.
2. Founders you meet once and feel they’re ‘different.’
3. Everyone is genuinely lovely, generous with their time and knowledge, and fun to be around.”

Top-shelf talent, great founders, and perhaps a cheeky pint after work. What more could you ask for?

Should you consider joining Granola?

If you have read this spotlight and feel like Granola is the right company and right team to build the primary interface for knowledge work, then the basic filter is: Do you live in London or San Francisco? If you don’t, are you willing and able to go live in London or SF? Perhaps you could get hired as a contractor somewhere else, but the entire full-time team at Granola works in-person.

Beyond that, “if acting with high caution or constantly trying to play devil’s advocate… are your default modes, Granola probably isn’t the place for you.” You should be the sort of person who assumes things can work, approaches your job with optimism, and can take a lot of autonomy.

If you want to increase the chances that your application gets looked at, you could put in a little more effort than usual. “We love to see side projects, or signs to show you care,” co-founder Sam said. (As it happens, we recently wrote an essay with some advice on doing just that.)

Most of the people I talked to described having a one- or two-week work trial in the office. “I felt I was able to have impact quickly,” Vicky, CX lead, said of her work trial. “Thankfully they felt the same!”

A few of the folks I spoke to told me that they had been rejected for jobs at Granola they’d originally applied for, but that the team had then reached out in the future about other roles. Another encouraging sign that—if the company seems interesting—you should let them know how you can add value.

Granola has a long way to go if they want to become the interface where most knowledge work gets done. But, with AI notetaking, they have already won the first game. Why not go help them win the next?

Thank you to Granola for supporting Next Play and making this spotlight possible.

Seven years ago, Brandon Hill’s parents came home with a stack of wholesale catalogues; thick papery booklets listing every product a grocery store could order.

“Wow,” Brandon said, curious, “Is this how they used to place orders back in the day?”

“No,” his father replied. “This is what they do TODAY.”

So began Brandon’s (and then co-founders Tre Kirkman’s and Robert Pinkerton’s) journey towards founding Vori: the startup building a modern operating system for grocery stores. Because although not every grocery store is still placing orders via paper catalogues, existing software is hardly better. “The grocery industry is one of the most under-digitized industries on earth,” Brandon said.

It is a nonobvious place to build. Maybe today is the first time you are reading the phrase ‘grocery store startup’. But the problem is real, the market is massive, and Vori already has encouraging traction:

• Processed $350M+ in payment volume in 2025

• Revenue scaling 5x year-over-year

• Generated $22M+ in incremental sales for independent grocers in 2025

• 577K+ supplier cost changes auto-detected by Vori AI, protecting retail margins

• Launched in 55 new cities in 2025, now operating coast-to-coast

• Customers (incl. Mollie Stone’s Markets) have become investors

(Go watch this video if you want to see a visual recap of Vori’s progress so far.)

Most startups you hear about are helping theoretical people you will never be on theoretical problems you will never experience yourself. Vori, on the other hand, is aiming to transform something that every single person reading this is familiar with: going to the grocery store.

What does the market look like? What is Vori’s plan for success? And should you join now?

Arming the rebels

Running a grocery store is a harder job than you might think. If you are a company like Amazon (which owns Whole Foods) or Walmart, you can smooth over some of that difficulty with billions of dollars that you use to build your own back-office teams and develop your own internal software.

But if you are an independent grocer—not a national behemoth—then you are in a tricky situation. You have to do all the same things Amazon does (ordering, sales, accounting, inventory management, etc.), and you have to compete with the big boys, but you have only a fraction of the resources.

So what do you do? Well, in the past, the best answer might have been: “suffer.”

Growing up, Brandon Hill (Vori’s CEO) watched both of his parents in the industry deal with systems that were built in the ‘90s and probably should have stayed there. If you run a grocery store today, there is no great operating system you can use. There is no Stripe, no Shopify. You are out there on your own.

And being out on your own is not a recipe for success. Especially when, as Brandon told me, “wages have stagnated, margins have compressed, and scale retailers [like Walmart] wield a significant pricing and technology advantage. At the same time, the industry is undergoing a massive intergenerational transfer of ownership: thousands of owners are handing the keys to a next generation that doesn’t want to run a business on paper, fax machines, and tribal knowledge.”

You often hear startup founders talk about ‘underserved industries’ that are ‘ripe for innovation.’ When you see ten different startups in the same industry building same-y products for that ‘underserved’ market, though, you start to wonder: is it really underserved? Or is this just to farm goodwill with investors?

Here with grocery, it feels as though Brandon, Tre, and Robert have encountered a market that is the kind of underserved, prime-for-disruption market that so many startup founders say they are attacking.

And starting with the $350B independent grocery industry is a carefully-calculated decision.

Bear hugs in the aisle

If Vori is successful, co-founder Tre Kirkman told me, the company will become “the clearing house for the global food supply chain. Every basket from Oakland to Auckland will be powered by Vori.”

That’s ambitious. But you have to start somewhere, and Vori started where it made the most sense: by building reordering software for independent grocers. Reordering software because it is one of the more tedious (and important!) things a grocer needs to do; independent grocers because they have fewer embedded systems and lower inertia preventing a switch to new software. (Also because, as it has become clear, the Vori team has a real passion for great independent grocers.)

Over time, Vori has built out their software into a full operating system for grocery stores: POS, payments, back office, price changes, and more; an operating system that has now facilitated more than $350M in payments and generated more than $22M in incremental sales for independent grocers in 2025.

Vori is an AI-native OS; the company is building solutions (actually useful ones!) with AI to let stores run on “full self-driving mode,” Brandon said. This is not a requirement—stores are not forced to give up the reins to AI—but the products are effective. To date, more than 577,000 supplier cost changes have been auto-detected by AI to protect retail margins for grocers. This is perhaps one of the few examples of AI tools in the market today where there is an obvious line between using the AI and increasing profit.

Then there are the intangibles. “Jaime, the dairy clerk at the Market at Edgewood, gave [co-founder] Rob a massive bear hug in the dairy aisle where he gave [Vori] a test drive. ‘I have goose bumps,’ he told us. This is when we realized the potential for life-changing impact,” Brandon said.

“The founders are close-knit and supportive of one another,” Klaire Korver, growth marketing lead, told me. All three are Stanford graduates. Brandon and Tre launched YC-backed social app Greo in 2017; Robert was building rockets at SpaceX and working on Lyft’s first fleet of self-driving cars.

Move fast, don’t break things

I do not think I have talked to a more grocery-obsessed group of people than the team at Vori. This is a group of people that cares deeply about grocery stores, about what they mean, about the value of independent grocery stores, about the problems grocery store owners face daily.

Vori is “customer-centric in a very literal way. Slack is full of customer conversation recaps — install learnings, post-live feedback. These stories directly fuel what we prioritize and what we build next.”

If you can’t empathize with the plight of the independent grocery store, well, I’m not sure if Vori will be a fit for you. You need to love (or be able to love) grocery stores and their importance. They are the customer!

“One of our core values,” Engineering Manager Clinton Blackburn told me, “is don’t mess with the dairy order. You mess that up and you’ve ruined a shopper’s visit to the store and potentially lost their trust. We move fast, but don’t break things because we know that the things we might break can ruin a business and someone’s livelihood. We keep this in mind with everything we do.”

It stood out to me just how emphatic people at Vori were about the no-politics culture. “I’ve worked in tech for 13 years across seed through public companies… And Vori is the least political company that I’ve ever worked at. Everyone’s ideas are valued,” Klaire Korver said.

“I don’t feel that there are any company politics to deal with here,” someone else told me. “Everyone is approachable and motivated and there’s no sense of work being above or below someone’s paygrade.”

This low-drama approach to things means that anyone—even a brand new junior hire—can make a big impact at Vori. “Titles aren’t a big thing here,” someone said. “Flat structure / real ownership. Everyone has meaningful scope, and that’s what motivates people.”

People work the hours they need to; some people told me they worked 8 or 9 per day, while Clinton Blackburn told me he works 12. “Did I mention I need more engineers to spread the workload?”

Should you join Vori?

We’ve established that you should care about grocery stores. But that’s easy enough to do by simply paying more attention the next time you shop. People 150 years ago had none of this.

Why else should you consider joining?

Jesse Lopez, Head of Product Marketing, had thoughts: “Over my career I’ve come to learn that what you do is really only 15-20% of what makes for an enjoyable job. Certainly, you want your work to be engaging and impactful, but in reality, who you work with, and how you work together, contributes way more to your enjoyment of a role. Now that we’re expanding our team, we often talk about carving out the most interesting work for our new joiners. Better to give the exciting new projects to the new hires so they can be easily engaged while they build rapport with their teammates.”

People who do best at Vori tend to be “highly curious, high ownership, customer-obsessed, and low ego. The last one is really important. We like to live by the proverb ‘If you want to go fast go alone, if you want to go far, go together.”

There is also the equity and growth trajectory perspective. Vori is growing fast, and it would not be surprising to see a Series B sooner than later (and then a C not long after that, if all goes well). This is a stage at which you can still get pretty meaningful equity (that could end up being very valuable if Vori ends up succeeding long-term). This is also a stage where you can make a big impact—your individual decisions could help shape the company’s direction—and be part of the reason Vori wins.

Vori is hiring across engineering, sales, marketing, design—just about everything. If the company sounds interesting, give them a closer look. And if you do reach out, you could try being unconventional: “We are down,” Brandon told me. “Shows creative hustle.”

Thank you to Vori for supporting Next Play and making this Spotlight possible.

Here is a seemingly basic question: What allows most modern war technology to work? What allows the drones, the autonomous vehicles, the GPS-guided missiles, the communication systems—and frankly, almost everything useful on a battlefield—to function properly?

The answer is the thing that the startup we are spotlighting today, CX2, is focused on: the electromagnetic spectrum.

Over the past few decades, billions of dollars have been poured into funding new defense startups and building advanced tech that is useful on the battlefield. This is good. Yet it is also true that almost none of these things work without connectivity. Disrupt that and everything stops working. Which is why electronic warfare has quietly become the most important thing to solve if you want to win.

“The electromagnetic spectrum is the base of the 21st century defense tech stack,” Porter Smith, CX2 co-founder and ex-Army commander, said. “If we don’t dominate it, none of the other fancy stuff works.”

CX2 is building the technology the West needs to dominate that spectrum. And I have a hard time imagining a better founding team for the job:

• CEO Nathan Mintz and President Mark Trefgarne are both former founders with successful exits.

• Lee Thompson was a SpaceX team lead for RF engineering.

• Porter Smith was a helicopter pilot in the Army for a decade and invested for a16z before joining.

The strong level of talent and military experience extends beyond the founding team, too: a third of the team at CX2 are veterans.

Better yet—if you are someone looking to make a big impact—CX2 is early. They raised a $31M Series A in 2025 led by Point72 ventures (participation from a16z, 8VC, Pax Venture, and others). They are currently hiring for their relatively small team at their El Segundo office; this is a unique opportunity for people, especially engineers, who want to work on hard problems with real stakes.

Over the rest of this essay I will share a little more from my conversations with the CX2 team. What is the market here? What are they building? And what would it be like to work there?

Invisible warfare

In the summer of 2024, Porter Smith was with troops in Ukraine when he noticed something odd.

“Ukrainian infantry guys take spectrum analyzers [devices to analyze the RF spectrum] with them everywhere,” he said.

Why were the Ukrainian troops doing this? Because ground forces were dealing with Russian jammers; devices that interfere with the RF spectrum. These jammers “affect your ability to maneuver on the battlefield. Because if you have large jammers, you don’t have GPS-guided munitions, you don’t have the ability to fly your own ISR assets, you don’t have the ability to fly your own logistics efforts.”

This obsessive focus on the RF spectrum is not the way war has always worked.

“Historically, [what the Ukrainians were doing] is a very niche electronic warfare task in the United States. And their version of 11 Bravos, standard infantry guys, were doing this.”

If the RF spectrum is upstream of most modern battlefield leverage, well, it is only natural that your adversary will want to disrupt it. This reality means that winning on the invisible, electronic battlefield is now the key to winning on the physical one.

So how would you go about this—practically speaking?

CX2 describes their approach simply: “Hunt the archer, not the arrow.” The thesis goes that the best way to dominate the RF spectrum is to identify enemy emitters (like jammers, or drone pilots) and engage them directly. “We think the 21st century kill chain begins in the RF spectrum,” Porter Smith said.

Hunting the archer

So far, CX2 has launched two products: Vadris and Wraith.

Both of them identify RF signatures on the battlefield. But they have different purposes.

Vadris is a small payload you strap onto an FPV drone that identifies adversary drone operators so you can engage the operator directly—not just their drones. “The counter-UAS market is flooded with drone intercept approaches,” Lee Thompson, co-founder and Head of Engineering, said. “We have flipped that on its head to instead hunt for the drone operator.”

Wraith is an autonomous recon drone that creates a heatmap of the RF spectrum and geolocates hostile emitters (like jammers and radars), even in GPS-denied environments. So you can dominate the spectrum even when connectivity is under attack (which is now the norm). If the 21st century kill chain begins in the RF spectrum, as CX2’s Porter Smith believes, then “Wraith is the spark for that kill chain.”

These two products are effective expressions of the same idea: start treating the electromagnetic spectrum as the battlefield itself. Because you need to, if you want anything else to work.

As you might expect, building these products has required a skilled team with a lot of drive and a lot of experience—you can’t really spend a weekend with ChatGPT vibe coding an instrument of war.

(Author’s note: I want to mention that while CX2’s market is currently focused on providing military technology to the West, there are other future applications for this too. Mapping and dominating the RF spectrum is useful if you want to keep drones away from arenas and airports, as one example.)

The team for the job

The team at CX2 is not your typical team; the environment is not your typical environment.

Usually, when you hear about impressive coworkers at your company, you hear people say things like “they were early at Stripe.” At CX2, you may be slightly more likely to hear “they were a commander in the U.S. Army,” or “they built RF systems at SpaceX.” The team at CX2 has a combined experience of more than 22 tours of experience in war zones. They know the mission because they’ve lived it.

As for the environment, “if you prefer to stay behind a desk… and you don’t enjoying being outside,” someone said, then you probably wouldn’t fit in at CX2. They test their products frequently in the Mojave Desert, at military bases, and at test ranges. In total, the team was outside testing for at least half of last year. “And 2026 will likely be even more than that,” I was told.

The culture at CX2 is rather intense. The stakes and standards are higher than at the average startup. The problems are more complex. “We have a few core values,” Nathan Mintz (CEO) said. “Show don’t tell; always remember who the end customer is; and to be open and transparent with feedback.”

Porter Smith described working at CX2 like driving in Formula 1. “Quick decision-making, rapid prototyping, and swift action are our DNA. An honest mistake caused by a sense of urgency is better than lethargic indecision or action. We prioritize and focus ruthlessly,” he said.

I do not get the sense there is much room for bullshit, or for puttering around, at CX2. And my favorite quote about the way they view work is just three words: “Get shit done.”

Nathan’s goal as leader is not to burn people out (“we take vacations and believe in weekends,” he said), but the “concept of work/life balance is certainly more biased towards a fast-paced startup environment. [This is not] for someone who is just looking for an 8-5 job.”

Amid all of this work on what is self-evidently an ambitious and important mission, however, I do get the sense that the team at CX2 is having fun. The company’s office is just a 7-minute drive from the beach, where one employee at the company says he surfs most mornings before work.

Responses about favorite stories, moments, and non-work pieces were intriguing:

• “Polish train rides, nights in the desert, and Drone Jesus.”

• “Going out to Lucerne to film our first product videos a month into starting.”

• “Our team spends a lot of time in the Mojave desert doing testing and we’ve developed a certain set of rituals and terms for the various things we see there. One time, one of our Airbnbs was supposedly haunted and someone found some sketchy letters from one of the previous tenants in a public drawer. It’s become part of our recurring lore.”

• “Don’t forget how close we are to the beaches of South Bay!”

Should you consider joining CX2?

CX2 is not your typical startup, and working at CX2 is not your typical startup job experience. And I think that, for the right person, CX2 could be an excellent fit.

Here is how I might describe the right kind of person:

• You want to solve highly technical problems; RF is a difficult domain and adversaries are working on their own solutions, too

• You find defense and systems engineering fascinating

• You want to build products that ship to real-world environments where the stakes are high

• You want to be outside testing products often

• The idea of defining how 21-st century warfare is fought sounds interesting to you

• You want to have a big impact with lots of ownership at an early-stage company

You could also ask yourself this question, which Nathan Mintz shared with me: “Are you interested in joining a top-shelf team that is reshaping how the West fights electronic warfare?

If this sounds like a problem worth working on, and a team worth working with, you can view open roles at CX2 here. The company is hiring at their El Segundo office and, right now, is mostly looking for engineers (both software and hardware).

Thanks to CX2 for supporting Next Play and making this spotlight possible.

2025 was a strange year for the public perception of AI-generated video.

As the technology to produce realistic-looking video on command (e.g. Sora 2 and Nano Banana) became widespread, so did the number of startups aiming to capitalize on this new capability. This was fun at first, but fatigue set in quickly. Really, another AI video generation app/website/product?

The word “slop,” now mostly used to refer to meaningless AI-generated content, was Merriam-Webster’s word of the year in 2025. Google Search interest for ‘AI video generator’ peaked in the fall of 2025 but has since begun to wane. Not everyone wants to see another wacky AI video of…Whatever.

There is a twist, though, there must be a twist, or we would not be writing this essay.

Amidst the graveyard of already-failed and currently-failing and soon-to-fail AI video generation applications there is at least one company that is doing better than ever: Higgsfield.

“On fire” would be an understatement to describe Higgsfield’s arc over the past year:

• Higgsfield publicly launched its video generation tools on March 31st, 2025.

• Since then, they have gathered more than 15M users.

• They currently report $200M run rate (doubled from $100M just 2 months ago).

And, in perhaps the biggest news, Higgsfield last week announced their $130M Series A funding round at a $1.3B valuation with participation from Accel, AI Capital Partners (Alpha Intelligence Capital’s US-based fund), and Menlo Ventures and others. Yes, those numbers are correct. For context, Anysphere (Cursor) raised $60M on a $400M valuation for their Series A; Anthropic raised $124M on $461M.

Higgsfield’s latest round is much like their growth over the past year—historic.

If you join Higgsfield to work under founders Alex Mashrabov, Yerzat Dulat, Mahi de Silva & co., you are not joining because you think this early-stage startup may one day become a unicorn. It already is. No, you are joining Higgsfield because you believe this story ends with Higgsfield growing to $10B, or $50B, or $100B, or more. You are joining because you believe Higgsfield can become one of the most important tech startups of the 2020s. Whether that will happen is anyone’s guess, but I hope the following piece can explain a little about what Higgsfield is doing, why they are winning, and whether you should join.

Not all AI video is created equal

Just five years ago, the idea that you might be able to generate a beautiful video with AI sounded like science fiction. It would have been outlandish to tell a social media marketer that they would be able to create video ads, user-generated content (UGC), and organic content, by typing a prompt into a computer.

In 2026, however, this is the reality of video generation. Models like OpenAI’s Sora 2 and Google’s Nano Banana have gone viral; anyone can now generate a video of almost anything. This should be exciting. It felt exciting at first to me, and probably to most people who tried creating their own videos.

But as I spent more time with the videos people were creating, doubt crept into the back of my mind. I had one thought: How much of this is actually valuable? A video of flaming giraffes jumping off a train Mission Impossible-style is cool, I guess. But what is the point? Where is the value?

What is the market here?

If the first step to AI video is simply having the raw energy—the magic—accessible, then the second step is figuring out a way to harness that magic. This is, as I learned quickly when talking to Alex Mashrabov, co-founder and CEO of Higgsfield, exactly what Higgsfield has done.

Most of Higgsfield’s run-up to $200M and 15M users thus far has been driven by marketing teams—especially on social media. “In the near future, half of all social media video will be generated with AI, and most of that AI-generated content is going to be created with Higgsfield,” Alex said. Jeff Herbst, an investor, described the social media market as bigger than Hollywood (another potential marker for Higgsfield).

To Alex and his team, the timing feels perfect. “First, generative AI has moved from being a research experiment to a practical, mass-market technology,” he said. “Second, the demand side exploded. Social platforms today require constant content. We are not trying to invent artificial demand; the problem already exists at a massive scale. Third, we are in the middle of a very strong global trend. Venture-backed startups succeed when they ride a wave, not when they fight against the current. And small teams [like Higgsfield] suddenly have an unfair advantage [because of] AI-native tools and workflows.”

Social media has (primarily) fueled Higgsfield’s ascent, but the potential market is much bigger than that. You could imagine a world where Higgsfield is the default creative tool for anyone who wants to create any video or create any visual effects. This is a market in the hundreds of billions annually, if not bigger.

“In 10 or 20 years, we see a company with billions of users, hundreds of millions of commercial videos, a thriving professional ecosystem, and billion-dollar revenues. A true platform of the industry, setting the standard for the industry,” Alex said.

Co-founder Yerzat Dulat is more laconic: “[we are] building the biggest media company ever.”

Building the biggest media company ever

Only a month after Higgsfield’s public launch, software engineers had a meltdown on Slack.

“I think we’re under a massive DDoS attack,” frantic messages read. “Servers are down. Everything is down. Traffic just spiked 20x.” Engineers worked through the night, calling AWS and Cloudflare, attempting to understand what was going on. Could it be real, they wondered? But DevOps experts all had the same opinion: “This cannot be real traffic. Nobody grows this fast.”

“And then,” Yerzat said. “Cloudflare confirmed. It wasn’t an attack. It was real people. Millions of real users trying to use our product at the same time.”

Thus marked the beginning of an eventful growth trajectory for Higgsfield that has not slowed down since. After launching their generative video tools, Higgsfield has become one of the fastest-growing startups in the world. The platform has attracted the attention of prominent cultural figures, including Madonna, Snoop Dogg, and Will Smith. As videos went viral, users poured in. Topline went up and up and up, and is now at $200M run-rate.

How? Well, you don’t need to look further than the team.

Alex Mashrabov, CEO, sold his former company AI Factory to Snap for $166M and went on to lead Generative AI at Snap. Yerzat Dulat, CTO, is a top AI researcher and author of popular open-source libraries on GitHub. Mahi de Silva, CSO, has decades of leadership experience (Triller, Amplify, Verisign). He was also a senior engineer at Apple in the late ‘80s and early ‘90s—how often do you hear that?

Look down the resumes of anyone else at Higgsfield and your reaction may be similar: impressive.

“Honestly, this is the most concentrated pool of top-tier talent I have ever seen in the market,” said Dalida Yerkuliyeva, Higgsfield’s VP of Talent. That is certainly the kind of thing you want to hear from someone like Dalida, who has spent most of her career helping startups hire great people.

“What first got me interested in the company was its fast growth - I have never seen anything like it,” Ruchita Gupta, VP of Finance, said. “PMF, traction, team are all unreal. This is a once-in-a-lifetime company and opportunity.”

To start, hiring was mostly in-house. “The first 50 people we brought on board were colleagues and experts that Yerzat, Mahi and I knew personally and trusted deeply,” Alex said. Higgsfield’s team has since expanded to well over 100 people. But, when I talked with the folks about the working culture, it seems like that early founding team environment still lives on today.

How hard can you work?

If I asked you to describe the “classic startup work experience,” you might describe a few things:

• Intense, long days

• A bunch of smart people all deeply invested in the problem

• Constant pressure, shifts, wins, and losses

• An expectation to ship extremely fast

• Huddled around laptops in a low-rent garage somewhere

Higgsfield doesn’t work in a garage, but the rest of it is basically accurate.

When new people join, Dalida told me, “you can see the hesitation in their eyes. These are talented people who [are giving up] a lot: comfort, family time. But the moment they see the team in action, working together, everything changes. You can feel the energy, the ambition… and it’s so contagious.”

“I think many outsiders assume that once a startup reaches our size, it becomes more corporate and comfortable,” Alex said. “In our case, it’s the opposite: we scaled the team, but we kept the founder mentality. At most startups the extreme ‘grinder mode’ usually only applies to the very first people - the founding team. We are no longer a tiny startup, and yet we still operate with that same mindset.”

Kamazhay Kospan, the marketing lead at Higgsfield, told me that although the work is intense, “it is hard to find the right word for [the founders’] remarkable trust in us. They really let us own our work and learn along the way. I’ve always thought: wow, Yerzat, Alex, & Mahi really believe in us and listen to us.”

“The magic of Higgsfield is our ability to execute quickly and thoughtfully,” said Matt Stratton, Higgsfield’s General Counsel. “Where some companies spin their wheels in an uncertain environment, we find a way to move forward.”

“Even when I give 200%,” she said, “I feel I can still do more. I don’t get burned out.”

Expectations are high, and they are consistent. “Yesterday’s success is just yesterday’s success. You can be extremely smart and skilled, but if you don’t keep delivering every single day, it’s easy to feel out of sync,” Dalida said. “[The intensity at Higgsfield] isn’t for everyone.”

So who’s a good fit? “Someone who doesn’t sit and wait for instructions,” Kamazhay said. “People who catch questions that are floating in the air, and who don’t believe that something is impossible. You need to be the type who dives in, figures things out, and trusts your ability to solve problems.”

If you “don’t need to wait for things to be fully defined before acting,” Jasmyn Jarnigan, VP of B2B Strategic Parternships, said, “and enjoy operating at the intersection of creative intuition and technical systems,” then you might be a good fit.

“The company is very flat,” Ruchita Gupta said. “If you want a fancy title or to avoid doing ‘analyst’ tasks, then this is definitely not the place for you.” Everyone is expected to do real work.

It helps to contextualize this within the reality that Higgsfield is one of the fastest-growing startups on the planet. The people who work there are talented and smart and believe they are building a generational company; they have already reached some historic milestones. If you are this kind of person, and perhaps you are, why wouldn’t you want to work hard?

Should you consider applying to Higgsfield?

There is one expected path for startups at Higgsfield’s stage: a modest $20M Series A, a few million dollars in ARR, and a few hard years of work to hopefully one day reach unicorn status and beyond. This approach sometimes works. Loom, for example—a name that most people in tech know and respect—peaked when it raised a Series C at $1.5B; this was viewed as a huge win.

With Higgsfield I think you have something of a different path. Something closer to a Cursor or an Anthropic or a Perplexity. You have an insanely fast-growing company that is raising way more money than companies usually raise; that is valued far higher than most companies at their stage are; that, if successful, could be one of the most valuable companies of its generation.

The other reason you join, I think, is because you crave something that few startups today can provide: an exhilarating startup experience alongside a lot of talented people. You want to work a 10-12-hour day when necessary. You want to be up late at night, solving problems together as a team. You want to become better at your craft.

And you want to be early at a company which could, if it completes its ambitious and admittedly difficult mission, become one of the most successful startups of the century.

As for how to apply? You can look for careers at Higgsfield here; the company is based in the San Francisco bay area and they’re hiring for both remote and in-person roles. They also have a presence in Kazakhstan. I asked Alex how he feels about cold emails; he’s open to them, though prefers thoughtful LinkedIn messages. (People often reach out with ideas that Higgsfield is already trying, he said, but you are welcome to reach out. Cold outreach is one of the best ways for candidates to get across Alex’s desk.)

You may expect a quick interview process (Dalida said it took “less than 10 days from my first interview to receiving an offer”) and onboarding is “extremely short. You don’t wait to learn everything first.”

If Higgsfield sounds like the kind of place for you, well, I’d encourage you to apply and/or reach out.

Thanks to Higgsfield for supporting Next Play and making this Spotlight possible.

TL;DR

If you only have a minute or two, here’s what you should know about Northslope.

• Built on a $450B OS. Northslope builds mission-specific AI applications on top of Palantir’s operating system. The team is led by ex-Palantir engineers who helped build that OS. They’re funded and led by former Palantir executives and recently became Palantir’s first (and only) Vanguard: Elite partner.

• Solving problems that actually matter. Some startups are in search of a problem. Fighting for product-market fit. Northslope has the opposite problem: there are near-infinite problems available for them to solve—problems like building rockets and delivering higher quality healthcare.

• Leaning into the FDE model. The vast majority of Northslope’s team has the job title Forward-Deployed Engineer (FDE). If you’re unfamiliar, these are people who embed themselves directly into customer companies to make purpose-built software for specific outcomes. This, Northslope’s founder Bill Ward says, is the way that quality software is made.

• Like a special ops team for software. Working at Northslope is a “choose your own adventure” experience. Your job description can’t predict exactly what you’ll work on because Northslope’s job is to solve, well, almost any problem that can be solved with software. The team has a flat hierarchy (their founder simply lists himself as an FDE on LinkedIn), and standards are high. If you join, expect to be surrounded by lots of smart people who will push you to do your best work.

Imagine there was a Fanduel-esque sportsbook for predicting startup success, and you were an oddsmaker. One day, a new file came across your desk. The facts were as follows:

• Builds real-world software for real-world problems on top of a ~450B operating system (OS).

• The team is made up of many people who originally helped build that OS.

• Started as a bootstrapped business, proving their profitable business model before raising any strategic capital.

• Found product-market fit fast, grew 6x in 2025, and has many large enterprise customers.

Chances are you would rate this company pretty favorably. You might even find these facts strange. What kind of early-stage startup runs a default profitable model? What “operating system” are we talking about? What does the company actually build? What do the people on the team do? How does any of this work?

I have just provided you with some facts about Northslope, a startup quite different from most of the other companies we write about at Next Play. Northslope is a team of engineers and operators, many of whom used to work at Palantir, who build applied AI software on top of Palantir’s OS.

Northslope is not a startup in search of a problem, desperate for product-market fit. No, the problems that Northslope’s customers face are both plentiful and important. Last year, they automated revenue management for one of the world’s largest solar operators. They helped build rockets. They helped a company process billions of dollars in weekly payments. And they solved quality metrics reporting for a major hospital, “taking people off paperwork and putting them on patient care.” I’m just naming a few.

Joining Northslope is like joining a special operatives team of sorts; a flat hierarchy with some of tech’s top talent who must be ready to go solve big software problems in the real world at a moment’s notice. Your job description cannot possibly say what you will be working on because Northslope’s problems are the world’s problems, and the world is an unpredictable place.

Curious about what it’s like to work at a company like Northslope, I sat down and asked the team some questions. How does the model work? What does collaboration look like? Who should consider joining?

A different way to solve problems

First, the model.

There is one approach to building software in which you create a rigid product with a bunch of general-purpose features. You ship that product out to the world and hope that people find some of it useful. Most software is made this way, and this approach can sometimes work very well.

Then there is the approach that Bill Ward, Northslope founder, describes: “The way you build high-impact software is by forward-deploying and building against specific missions and outcomes. You build for the people, business models, and data that make an organization special - bespoke software that compounds their competitive advantage, instead of commoditizing it.”

This second approach, which is usually called the Forward-Deployed Engineer (FDE) model, is what Palantir pioneered in the mid-2000s. You build a core platform (like Palantir’s Gotham or Foundry) which is capable of doing a near-infinite number of things. Then, you embed your own engineers—your FDEs—into customers’ companies to help them build specific software applications for their problems.

In the early days of Northslope, Bill Ward & co. found themselves working with a massive retail chain that was struggling with tax reporting. No existing SaaS product worked, so they were paying millions per year to a big 4 consulting firm to take care of it, which was expensive and still led to mis-filed taxes and fines. Enter: Northslope.

“We sent two people for three months and built custom, mission-specific software that allowed their internal team to run the entire process themselves. They eliminated the high seven figure annual contract, integrated all their data, filed on time, could respond quickly to state disputes, and avoided penalties. All while dramatically reducing operational overhead.”

This story captures the essence of the FDE approach: it is all about delivering tangible outcomes that move the needle.

Instead of creating a piece of software that kind of works for a lot of companies, you create lots of different pieces of software that work really well for specific companies. The commonality is that they are all built on the same platform; the same operating system.

The market for building on Palantir’s operating system is huge. As of this writing, Palantir is worth approximately $450B. Their customers include everyone from Wendy’s, where they make sure the company doesn’t run out of Frosties, to the United States government.

But although Palantir popularized the FDE model, they are not a consulting firm. “This was hard for a lot of people to understand when I worked at Palantir,” Bill said. And, still, there are many companies with important problems who need help building on top of Palantir’s foundations. So there is some tension:

Who is going to help the 1,000s of companies that need purpose-built AI applications on Palantir’s OS?

Enter: Northslope. A team of 95+ people, mostly FDEs, many of whom helped build Palantir itself. A team ready to tackle some of the world’s biggest problems at the drop of a hat.

The everything company

One year after founding Northslope, Bill Ward posted the following on his LinkedIn:

It is a rare few startups that get to make such a large impact—and on so many problems—so quickly. But this was not new terrain for Bill, nor for many of the ex-Palantir engineers (like Nathan Cole and Michael Gamble) that he hired.

“Northslope’s rapid growth has been impressive and puts them in a category of their own,” Palantir’s Global Head of Commercial Ted Mabrey said in a major announcement that Northslope would be the first company in Palantir’s new Vanguard: Elite partner category.

The nature of the work makes it both exciting and unpredictable. “This is a choose your own adventure company,” Northslope’s CCO Max Helzberg told me. “Come in, get exposure to different problems, and start to build your own practice; your own company within the company.”

I asked Bill what the FDE onboarding experience was like at Northslope: “Monday of week three, you’ll be on your first customer project. You’ll show up—hopefully in their office, maybe on zoom—and ask: ‘What is the hard problem you have?’ They’ll describe it, and you’ll start building software iteratively with them.” Within two to three days, Bill says, you’ll ship something, get it in front of a user, and then build the next version.

One other benefit of Northslope’s business is that it does not necessarily need to follow the traditional VC path. Though Bill Ward describes Northslope as a software company, not a consulting business, it has had revenue since day one, and earns as it builds.Their seed round helped to accelerate hiring and product development , but don’t mistake Northslope for the “who cares, let’s just raise every year and figure out profit later” sort of organization. Northslope raises money to aggressively invest in better serving more customers, in more places, not to survive and pivot.

Flat hierarchy, tall responsibility

If you want to build a truly great sports franchise—like the Bill Belichick-era Patriots—you need a few things. You need a lot of great talent. You need high standards, both personally and collectively. You need a deep level of trust and respect between every single player. And you need players to recognize that the team wins as a team; there’s no hero ball. Everyone helps everyone else. The hierarchy is flat.

From what I can tell, Northslope seems to have all of these ingredients.

On the talent front, “Bill’s superpower is convincing incredibly talented individuals to work with him, figure out their superpower, and unblock and unleash them,” Max Helzberg said. Perhaps the best metric to demonstrate this success is that, at least at one point, the team was more than 90% staffed by Palantir alumni; people who, generally speaking, have plenty of options as to where they’d like to work.

“Individually, everyone is very competitive and holds themselves to high standards. However, they understand that we win as a team and it’s important to give feedback, share learnings, and help others along the way,” said one of Northslope’s recruiters.

“The org is supremely flat. Everyone is an equal. Coworkers are friends. Everyone cares,” someone else told me. And there is perhaps no better, albeit superficial, evidence of this than that Bill Ward, who founded the company, lists himself on LinkedIn as a Forward-Deployed Engineer. Just like any other engineer. In some contexts this could seem performative—is the hierarchy really flat?—but at Northslope a flat hierarchy is the natural default. For a company like theirs, it is the most effective way to work.

High standards mean high trust, too. “The level of trust awarded to all Northslopers from day 1 was the biggest appeal to me. Everyone is brought here because of a distinct skillset that they bring, and they are immediately trusted to use that skillset on implementations,” said another FDE.

It may not surprise you to learn that the team moves quickly. Nikhila Obbineni, an FDE who previously worked at BlackRock, was “looking for a place where I could wear multiple hats and move fast. I wanted to own things from the ground up rather than be a small cog in a massive machine. Here, I can go from ideation to implementation in days.”

As for the fun stuff? The company does twice yearly full-team offsites, the most recent of which was in Portugal. But plenty of companies do that. More unique were tidbits like: “The London office independently created a meme subculture revolving around frogs, while the NYC office independently created a meme subculture revolving around pigs.” Regarding the latter, I can confirm that everyone I talked to mentioned a game called Pass the Pigs. If you want to play, well, first you’ll need to apply.

Speaking of…

Should you join Northslope?

There are a lot of credible reasons why a smart and talented person would want to join Northslope. That is not something you can say about every startup in the world. One early employee explained it well:

“My fulfillment in a job boils down to three things, which are really hard to find together: smart and caring people, a business being built for the long-term, and teammates I’d lay down my professional life for. Northslope was the first company I found in a long, long time where all of these were true.”

One reason you may want to join is you’d like to join the startup equivalent of a high-performance sports team. You’d like to work somewhere where the standards are high. Where everyone cares a lot. A place where you can be surrounded by smart people (probably smarter than you!) every day.

Or perhaps you’d like to join Northslope because you want to put your skillset to work on real problems. Problems like, as we covered earlier, helping companies build rockets and deliver healthcare. Maybe you are bored of a job you do not think is particularly meaningful; maybe maintaining the same old codebase for years on end is not pushing you to your limits. If you want to use your talents in the real-world for products that will actually be used—not for a startup in search of a problem—consider Northslope.

You may also want to join because you share Bill Ward’s vision for Northslope, a vision for “all of the world’s most important institutions should be able to run like SpaceX - building their own software for their own specific missions We’re building the applications on the Palantir operating system that enable them to do that.”

This vision alone, if you believe Northslope can achieve it, is reason enough.

But who are Northslopians, really?

“The people who really thrive here are the ones who get genuinely excited about untangling messy problems. Like when you finally crack that complex data pipeline at 10pm and you’re actually pumped about it. They’re intellectually curious to the point of obsession, spending lunch breaks refining solutions just because the current one bugs them,” Nikhila Obbineni said.

People who don’t fit in well are people who “want a detailed instruction manual,” and people who “want a clear ladder to climb based on a prescriptive list of milestone/achievements.” Northslope is a flat, choose-your-own adventure company; corporate ladder-climbing is a far cry from its culture.

If you want to join, Northslope is currently hiring across multiple departments (mostly engineering and operations) at their offices in London, Dubai, Denver, Washington D.C., Seattle, and New York. Fill out an application if something seems relevant, but don’t forget that you can also get more creative and you can consider sending a useful cold email to the right person at the company.

TL;DR

If you only have a minute or two, here’s what you should know about Hearth.

• Automating the entire revenue cycle for contractors. Roofers, plumbers, landscapers, and other contractors spend countless hours on sales and revenue work. Hearth is using AI to build a completely automatic revenue engine—contracts, invoices, estimates, financing, phone calls—so contractors can get back to focusing on what they do best.

• $35M+ ARR and turning a profit. Hearth has 15,000 paying customers and is profitable. Their new AI receptionist product, just one new product in the company’s current shift from “tool” to “ teammate,” hit $1M in ARR in a quarter with no advertising.

• Truth above all else. Most culture at Hearth is downstream of truth. Feedback is direct and honest. Priorities shift as fast as reality does. When measuring performance, Hearth cares about quantitative output above all else. Pursuing truth is how Hearth’s team wins.

• “High slope” over impressive resumes. Hearth is looking for true builders—people who are founders at heart. The best way to get hired is to show that you are good at what you do and capable of learning quickly. You’re welcome to send a cold email if you’d like to get hired. Misha, the CEO, prioritizes “show, don’t tell.”

Why most small business contractors fail

Nobody starts a roofing business so they can spend all of their time behind a computer:

• Responding to leads before competitors do

• Creating estimates that convert

• Wrangling financing products so they can close deals with flexible payment plans

• Sending invoices

• Chasing down those invoices when they don’t get paid

• Handling the 100s of other revenue management tasks that comes with running an SMB

No, they start a roofing business because they want to install roofs and make money.

The same can be said for plumbers. For landscapers. For carpenters. For people who install windows, doors, and insulation. For deck builders. This is true for all contractors in the trades: they would much prefer to spend time on their business instead of doing sales and revenue work with clunky software.

But in the past, the version of the world where contractors can just focus on the physical parts of their businesses has not existed. Contractors are shockingly underserved by software, stuck working with clunky tools and heaps of dashboards and CRMs that aren’t automatic enough.

Hearth is aiming to change all of that by using AI to automate the entire revenue cycle for contractors. The company was founded by Anthony Ghosn and Joe Lonsdale (Palantir co-founder), and since 2022 has been run by veteran CEO Misha Tsidulko. The signals to date are impressive:

• 15,000+ paying customers, from family-run businesses to large franchises.

• $35M+ ARR in 2025.

• Profitable (!)

• Has raised $61.8M in total from 8VC, Founders Fund, Human Capital, and others.

• 200 employees and growing fast.

This is some of the information that made me consider writing about Hearth; information that I thought might make talented people want to apply for a job there. Lots of revenue? Profitable? Strong fundraising from top venture capital firms? Co-founded by one of the people who started Palantir? Hmmm. Interesting.

Hearth is now hiring for lots of roles both remotely and in their New York City office. So I sat down with the team and asked them questions to figure out how Hearth gets work done, what makes them special, and what it is like to work there. By the end, I hope you will have an idea about whether you’d like to join.

Trades are the backbone of America, but bad software is killing them

Put yourself in the shoes of a roofing contractor for just a few moments. You have spent the last few years of your life honing your skills and now you are ready to go start out a venture of your own. This feels like the American dream: work a job, earn some skills, and start a business of your own. Congratulations.

It’s only after you open the business that you realize all is not well.

You spend all day on a jobsite with your team and get home at 10 P.M. to realize that you still have six hours’ worth of contracts and estimates to send out. Then half of those customers don’t pay you on time, so you have to spend other odd hours—mornings, nights, whenever—chasing down those invoices. Your friends & family referrals lull after the first few months, so you realize you need to start marketing, an entirely new task. That marketing produces leads, which you suddenly have to dedicate heaps of time towards ‘qualifying’ and ‘nurturing’ and ‘converting’. Oh, and remember, your job is to install roofs.

Doesn’t this all sound a bit ridiculous? Why can’t you just focus on what you are good at?

The answer is that there is not much good software to help you. “More than 50% of small business contractors who keep our homes running fail by year 5,” Misha Tsidulko, Hearth’s CEO, told me. “Even worse than restaurants.” When I asked about existing software, Misha said: “Contractors don’t want software. They don’t need another CRM or dashboard. They need revenue.”

The market for solving the revenue management problem is big; the kinds of companies Hearth serves generate more than ~$1T in annual revenue in the United States alone. If Hearth is able to become the default way that these companies acquire and manage revenue, then they are in position to become, as Misha describes it, one of the “great American technology companies.”

I often hear from people who do not care all that much about the problem their startup is solving. One thing I find encouraging about Hearth is that their team seems to be deeply invested in the human reality, and the pain, of this problem. “I come from a family of contractors and am married to one,” someone at Hearth told me. “The trades are the backbone of America in so many ways, and I wanted to be a part of the solution.” It was surprising to me how much of the team I talked to had real, personal ties to the trades and deep empathy for the problems that their customers face.

Empathy for the problem is one thing, though. Building a great product is something different altogether.

Racing from $0 to $35M to $350M—and beyond

Hearth began its life in 2016 as a product that let contractors offer customer financing to their customers. “Our solution democratized access to customer financing for small business pros, disrupting the old-school ‘dealer fee’ model which charged 10-20% of the job, which was a total scam,” Misha said.

Since then, Hearth has gained a massive amount of traction. The company currently has more than 15,000 paying SMBs, most of which landed on Hearth via an efficient acquisition engine that pays for itself within 30 days of a new customer signing. The product suite is much bigger now, too: digital estimates, invoices, payment processing, marketing tools, and a lightweight CRM have all been added over the years.

These products, along with the large customer base and the proprietary data Hearth has collected, have paved the way for a major shift: Hearth is now transitioning from a tool to an actual teammate by building an AI platform layer on their existing infrastructure.

“We are undergoing a new transformation that will allow us to realize the most ambitious version of our mission,” Misha said. “Hearth is becoming an intelligent, personalized ‘team member’ designed to autonomously drive revenue.”

The previous version of Hearth—the one that generated $35M in revenue last year—was a tool contractors could use to help them handle revenue management. The company that Hearth is becoming is one that will do the work for contractors end-to-end. Leads, estimates, invoicing, phone calls, financing, and more will all be handled by Hearth’s AI.

Perhaps the most impressive public step Hearth has taken towards building this AI layer so far is their newest product, AI receptionist, which launched in 2025. The product answers calls 24/7, books appointments, offers financing, captures lead, and will soon be able to go outbound and follow up on estimates. Customers are delighted and pushing for more.

Hearth’s AI receptionist hit $1M ARR in just three months and with no paid marketing. “Just paying customers and word of mouth,” Misha said. It’s encouraging traction for Hearth’s new direction, which is far more ambitious than merely an AI receptionist. Within a couple of years, Misha said, “Hearth will be the indispensable infrastructure that runs the revenue engine for $5M—$25M home services contractors.”

So far, Hearth has built a suite of useful tools for home service contractors. This was the path to 15,000+ customers and $35M in 2025 revenue. Misha believes that the path to $350M in revenue (and beyond) is to to leverage proprietary customer financing and workflow data, and deploy intelligent, personalized AI crew members who just do the work of the entire revenue cycle, helping pros win more jobs and get paid automatically.

You may be wondering: Why can’t any other AI startup just jump in and do the same?

It’s here where Hearth has (another) big advantage. The company has spent a decade in this space. With that time, they’ve accrued huge amounts of proprietary SMB and customer financing data. They’ve built an existing distribution base of current and old customers. And they’ve designed a sales motion that covers CAC within 30 days. A new startup trying to solve these same problems would have a decade’s worth of hurdles to jump first.

Building reliable AI agents for contractors means solving for complex technical problems at scale:

• High stakes (missed leads costs real money)

• Multi-system orchestration across payments and scheduling

• Users on roofs, not at desks

• Network-level learning across thousands of contractors

• Zero lead leakage across every channel

• Sub-second response, near-perfect accuracy

• Chaotic inputs (voicemails, missed calls, partial conversations)

Building a great product here requires a great team. So I chatted with Hearth’s employees to figure out what sort of culture has made all of this possible—and who they want to hire next.

Truth and competitive greatness

You would be surprised how hard it is to discover what it’s like to work at a startup. People default to platitudes. “We work hard,” someone will say. “It’s fast-paced.” The irony is that while these answers are often true, they tend to be useless if you want to know what it’s really like to work there.

Luckily, the team at Hearth opened up about some of the most unique facets of their culture.

One of the overarching virtues upstream of Hearth’s culture is “Truth”. I recognize that may sound fluffy. Truth? What does that mean? Wouldn’t every company say that?

Hearth’s team, though, seems uniquely obsessed with truth. It informs every single part of their culture. Let me give you a few examples of how this looks downstream:

• The company cares about quantitative performance more than almost anything else. Is your work producing obviously good outcomes? Can you clearly point to those outcomes? The people who can answer “yes” to these questions are the ones who thrive at Hearth. “There’s very little performative startup culture,” Misha told me. “What matters most is whether something works.” Strong outcomes are rewarded at Hearth.

• With truth comes a high level of trust, so people are treated like adults. If a friend shadowed me at Hearth for a week,” one employee told me, “they’d be surprised by how hands-off management is. My boss isn’t constantly checking in or asking for updates. You’re expected to know what you own and just run with it. No one’s micromanaging, but that also means you can’t hide or wait around for direction.” Amanda Hester (Director of CS), told me that “Hearth trusts people to act like adults.”

• Communication is honest and direct. Nearly everyone I talked to emphasized how direct the communication is at Hearth. (We do a lot of spotlights like this at Next Play, and no group of interviewees to date has emphasized direct communication as much as Hearth’s employees did.) “People communicate very honestly with each other,” someone said. “It’s normal to be candid with leadership, disagree openly, or push back if you think something doesn’t make sense. There isn’t a lot of tiptoeing or ‘let me word this carefully so it lands right’ energy.” At Hearth, you can speak your mind.

The other thing that stood out to me was this idea of “Competitive Greatness”. Hearth’s team believes that greatness is a “conscious choice to be at your best when your best is needed.” And your best will be needed often at Hearth, a company that builds new products fast and is currently undergoing a significant shift as they build an automatic AI layer.

To achieve greatness at Hearth, one employee told me, “you need to be able to execute FAST. This is a constantly changing environment (changes for the better) because the company is constantly listening and can make pivots quickly. Other cultures I’ve worked in had levels and levels of red tape—not Hearth.”

“And when disagreements arise,” someone else said, “data wins. It’s based on data, not the person.”

One final tidbit: people stick around at Hearth for a long time. “Hearth has a lot of people who have been here for five or more years, which is unusual for a tech company that’s still in a startup phase,” someone told me. “It says a lot about the culture, how people are treated, and what keeps them around.”

Should you join Hearth?

There’s one term that Hearth employees—and their CEO—use to describe who fits in: high slope.

“We care far more about how quickly someone learns, improves, and takes on responsibility than where they came from or what their resume says,” Misha told me. People who thrive at Hearth are people who are hungry to become the best versions of themselves, who have a high velocity of learning and growth, and who want to do impactful work. Hearth rewards team members who are resourceful and create their own momentum.

Hearth is, in other words, looking for people who can actually do the work.

In particular, Hearth right now is looking for strong technical builders—from backend engineers to designers to senior PMs. They are also looking for a CTO-level leader to partner with Misha as Hearth builds out their autonomous revenue engine for the trades. If that’s you, reach out.

If you’re unsure, I’d ask you a few additional questions before joining:

• Can you empathize with contractors’ pain? Because the team cares a lot.

• Are you someone who cares a lot about results? Performance is measured by outcomes.

• Can you be flexible when there are disciplined pivots? Priorities change as reality does.

• Are you happy with direct, honest communication? This is the style at Hearth.

There is also the benefit that Hearth generates tens of millions per year in annual revenue and is profitable. There is generally less risk associated with working at a company like Hearth as compared to working at a company that relies on more fundraising rounds to avoid going bankrupt.

If this sounds like you and you want to apply, I’d encourage you to get creative. Misha told me that Hearth has found most of their employees via referrals, outbound, and investor networks. They’re looking for people who are “founders at heart” and who can take ownership. This isn’t to say that you should not apply to the open positions on their website, but you may want to consider doing more.

When I asked Misha if they are open to receiving cold outbound, I got a resounding yes. “We are open to thoughtful cold emails, prototypes, and other ways to ‘show don’t tell’ or add value. These are all stronger signals than a polished resume.” If you are not sure how to write a good cold email, read this essay.

And if you know you can add value but don’t think you have time to write some fancy creative cold email, “don’t overthink it,” Misha said. “Just drop me a note and say hello.”

You can browse all of Hearth’s open roles on their website. If you really want to join, I’d encourage you to reach out with something valuable even if your specific role isn’t listed on the website. Like with any startup, if you can help Hearth make more money, they may want to talk to you.

And finally, it’s useful to re-iterate: If Hearth succeeds, in just a couple of years they will be the default revenue engine for home service contractors. This is a large market. If you believe in the problem and the mission, and if the culture sounds like a fit, I’d encourage you to reach out.

Thanks to Hearth for supporting Next Play and making this piece possible.

TL;DR

• Product-market fit surprisingly fast. Fulcrum is a small, relatively young company. And yet they already count 25 percent of the top 100 insurance brokers in the United States, including M3 and Hylant, as customers.

• Solving the hardest problems first. Instead of starting with low-hanging fruit, Fulcrum tackled policy checking: a tedious, 8 to 10 step process with a ton of nuance that used to require lots of frustrated human labor. Fulcrum’s AI can now handle this, and other things, end-to-end. Customers have described Fulcrum as the “Ferrari of policy checking.”

• A valuable, but unsexy and underserved, industry. The top 100 brokers in the United States brought in $80.5B last year; the number is $260B for the whole U.S. market. It’s a gold mine of a market that is, by nature, obsessed with the fine print. But with that fine print comes a lot of work that is both tedious and important, which Fulcrum is automating with AI.

• Early-stage with massive upside. Fulcrum is small, already has enterprise PMF, and aims to become “the operational backbone of all commercial insurance.” It’s the kind of profile you might look for if you wanted to get in on the ground floor of something potentially huge. Fulcrum is hiring across most departments in SF and is looking for “completionist” people who go deep on nuance.

The conference

You can divide the lifetime of most startups into two spaces: the time during which the founders hoped the idea would work, and the time during which the founders knew the idea would work.

Fulcrum co-founders Arjun Mangla and Sambhav Anand entered the latter space at a conference for insurance brokers in 2024. “We showed up with what I would call half-built workflows, and yet we were still getting stopped constantly,” Arjun said. “Then brokers started calling Fulcrum the ‘Ferrari of policy checking’. People were backing up what they were saying by signing contracts with us at the conference.” That was when it was clear, Arjun says, “that we’d hit product-market fit and needed to double down and scale.”

And scale is precisely what Fulcrum has done. Their product, which automates many of the most important and tedious workflows for people who work at insurance brokers, is currently being used by nearly a third of the United States’ top 50 insurance brokers (including M3 and Hylant). They’re backed by South Park Commons, Broker Tech Ventures, and Foundation Capital, among others, and reached product-market fit (at large enterprises!) while they were still a team of just 12 people.

Fulcrum is currently scaling fast and is hiring across most departments. This could be one of those unique opportunities to get in on a rapidly-growing startup that’s found product-market fit during its earliest innings. If Fulcrum wins big, joining now would mean that you have a good shot at winning big, too.

I wanted to learn a bit more about Fulcrum: what they’re building, where this thing is headed, and what kinds of people would be a good fit to join. I chatted with the team and, below, will share what I learned.

A gold mine with structural problems

Imagine a probably-real woman named Kimberly. She is an Account Manager at one of the largest insurance brokers in the United States. She works in a modest office building in Des Moines, is paid relatively well, and only has to work extra hours sometimes. From an outsiders’ perspective, there is not much complaining you could do about her job. But to Kimberly, some of the work is a nightmare.

One thing Kimberly has to do, for instance, is something called a “policy check”. This is a rigorous review of the insurance policy that has been offered to a client. The goal is to make sure that the insurance policy matches what the client has been quoted; in other words, to verify that the client was not promised something that does not appear in their insurance, which can cause a lot of problems down the road.

To do this well, Kimberly has to pore over the often-gigantic insurance policy with a fine-toothed comb to make sure everything lines up. This takes many hours and is horribly tedious. It is also not a good use of Kimberly’s time. Running a policy check is not the most skilled work she could be doing.

Policy checks are just one of the tedious tasks that Kimberly has to do, but does not particularly want to do: quote comparisons, proposals, and client prep are some of the others. If Kimberly had a magic wand to wave, she would materialize a full-time team of super talented junior employees that could do all of this work for her. This would allow her to work more effectively and manage a bigger book of business.

If Kimberly’s employer (the massive insurance broker) had a magic wand, they would no doubt do the same. The issue is that this is not really possible. No company can afford to have full-time teams of people who only do policy checks or client proposals. Instead, they run with the status quo: talented people do this tedious work in-house, or the company outsources to offshore companies that tend to be low quality and are not always very reliable. But what if that magic wand did exist?

If Fulcrum succeeds, their AI tools can be the solution to this—and other—problems in insurance.

The market, while somewhat unsexy, is a gold mine. Insurance brokers have a lot of money: the top 100 brokers in the United States made $80.5B in 2024, and the total USA market size is ~$260B. And insurance work is, by nature, steeped in fine print. With that fine print comes a lot of tedious work (like Kimberly’s policy checks) that Fulcrum has built a product to automate.

Solving the hardest problems first

There is one approach to startup-building where you start by attacking the low-hanging fruit. You do what is easy, pull in some initial customers, and then use that momentum to attack the hard stuff.

Fulcrum has taken the opposite approach by starting with some of the most nuanced work in insurance. “Policy checking is one of the most risk-sensitive flows in commercial insurance,” Heffernan Insurance Brokers, one of Fulcrum’s customers, wrote. “Even small discrepancies between [what is promised and what is on the policy] can introduce errors and omissions (E&O) exposure.”

But policy checking (and other workflows, like quote comparisons and proposal creation) are some of the most valuable work that insurance brokers would like to take off their team’s plates. So the team at Fulcrum decided to start there. “We always take an end-to-end view: if we touch a workflow, we own it all the way through,” Arjun, Fulcrum’s CEO, told me.

“Policy checking isn’t a cute one-step task,” Arjun said. “It’s an 8 to 10 step process involving hundreds of pages of PDFs, thousands of data points to reconcile, coordination across legacy systems, and tight coupling to the downstream policy delivery process.”

Building a useful version of Fulcrum, then, meant building an AI that would not hallucinate; an AI that was extremely smart; an AI could handle all of the nuance that humans need to handle.

It appears that Fulcrum has achieved such a feat. Today, nearly a third of the top 50 insurance brokers in the United States are Fulcrum customers, including M3 and Hylant. Customers are reporting that their employees are saving 6-8+ hours per week and can handle bigger books of business. Insurance brokers are able to scale further without adding headcount, which is something they are willing to pay a lot for.

Fulcrum has quietly accomplished this without any major funding rounds and without a huge team. I wondere—how? So I started asking questions about their culture.

Going the distance, fast

Things move fast at Fulcrum. “This is my seventh week at the company,” a founding member of the sales team told me, “but I feel like I have already been here for a year.”

Lexie Tonelli, who leads strategy and ops, said that she’s seen co-founder and CTO Sambhav “build in a single weekend what other people have entire companies built around. He’s someone our younger engineers really look up to and learn a ton from, and watching him work makes me feel very confident that we have a true 100x engineer leading the team.” High praise.

More details from the team gave me a similar impression:

• “Blink once and there’s a new product or feature released.”

• “Team meetings are incredibly productive and effective, [not just] a box to check.”

• “People who enjoy living in that tension between speed and completeness [do best here].”

There has been a lot of discussion this year about how many hours people should be expected to work at early-stage startups. When remote work surged during COVID, it felt like things were more relaxed at some companies; work-life balance was a hot topic, anyone could work from home, generous PTO policies were en vogue. Lately, there’s been something of a vibe shift—996 is real at some companies.

So I asked Fulcrum: What is the work expectation for employees?

“This isn’t a 996, but it’s also not a 9 to 5,” Lexie Tonelli told me. “We stay late because we take our work really seriously and are working closely together in person at our SF office.” Not all of this staying late is a relentless grind, though. On Fridays, “Sambhav [co-founder] orders croissants in the morning, and in the afternoon we have our weekly retro - an all-hands where people demo what they’ve been building and we align on priorities for the next week. We usually wrap up with a happy hour, which becomes a nice chance to properly catch up with everyone outside of day-to-day work.”

Should you join Fulcrum?

One of the main reasons you might join an early-stage startup, as opposed to a more established company, is for the equity. Cursor is one recent example—someone who joined after their seed round just 2 years ago is likely worth many millions on paper as of Cursor’s most recent funding round.

Fulcrum looks rather appealing on that front: they are early-stage, which means you likely get more equity than if you joined somewhere at Series C. They have already found product-market fit with some of the biggest insurance brokers in the country. And their vision is massive.

If Fulcrum succeeds, Arjun said, they “become the operational backbone of all commercial insurance. [Eventually], that same orchestration layer [will] extend beyond the brokerage to the entire ecosystem: risk managers, brokers, and carriers all running on a shared data and analytics platform.”

There is more than just the equity, of course. Another reason you might join is to work on a product that fundamentally makes people’s lives better. Kimberly, the possibly-real Account Manager we covered earlier in the essay, is genuinely happy that Fulcrum is taking all of this tedious work off her plate.

If you apply and Fulcrum takes interest, you can probably expect a work trial that resembles the kind of work you’ll do at the company. Throughout the process, you should get a good idea about who you’ll be working with and what they are like. “I had interviewed with other Series A-ish companies that I turned down,” someone on the Fulcrum team told me, “because I didn’t think the founding team was very compelling. Fulcrum was an exception. I felt like I got a really good sense of the founders.”

On the engineering side, you should “care about building really high-quality software.” Founding engineer Tristan Rhodes said that his goal is that Fulcrum’s engineering team is known for building “the most reliable and delightful software provider in the insurance brokerage space,” a goal that requires software engineers who are “honest, appreciate craft, and like to work hard.”

More broadly, you are more likely to succeed at Fulcrum if you are someone who:

• Doesn’t like “big tech bs”

• “Can appreciate craft, is honest, has a high standard for things”

• Is willing to “spend a significant amount of time learning about the insurance industry”

• “Care about building really high-quality software”

• Prefers “respectful honesty” as opposed to “lies to spare your feelings”

• “Have a high bias for action”

Sound like your kind of place? Fulcrum is currently hiring across most departments. They work in-person in San Francisco. If you do not see a fit, you can still try reaching out to someone at the company and explaining how you think you would make Fulcrum more valuable. If you go with that route, read our guide to getting a job with cold email so the message you send to Arjun (or whoever you email!) ends up as something they will find useful. (A LinkedIn DM could work too, one team member told me.)

Thanks to Fulcrum for supporting Next Play and making this piece possible.

TL;DR:

• Born from a $10k legal bill. Scott’s first startup paid out half of their early funding in legal fees. This frustration sparked Spellbook: a mission to make legal work 10x more efficient using AI. Now their product is used by lawyers at eBay, Nestlé, Crocs, and major law firms to draft and review contracts much faster.

• An eight-year overnight success. Founded in 2017 as a templating tool, Spellbook pivoted repeatedly before finding product-market fit in 2022 after GPT models emerged. The waitlist exploded to 30,000+ lawyers within weeks. They credit the Buddhist philosophy shoshin (“beginner’s mind”), which allowed them to stay open to new possibilities throughout the years.

• The StarCraft principle. In the video game StarCraft, speed wins and advantages compound. The same may be true about the real-life game of founding a startup. Spellbook moves “comically fast” with minimal meetings, async work, and a bias toward doing over planning.

• Building for a skeptical audience. Lawyers are uniquely dubious and detail-oriented, so shipping shoddy MVPs kills momentum. Success at Spellbook means building trust with customers that the product will always be at a high quality. Move fast and build trust, not move fast and break things.

• Impressive momentum. Spellbook raised a $50M Series B from Khosla Ventures back in October. More than 4,000 teams of lawyers, including at some of the biggest companies in the world, now use the product. Spellbook is currently hiring across almost every department.

The legal bill

Scott Stevenson’s first side project out of college was not an AI startup. It was an instrument.

The sci-fi-esque object, called Mune, was a new way to perform electronic music. Scott spent long nights designing the product, coding the interface, getting the little details right. After a while he launched a Kickstarter, and raised a $20k angel check. When he felt the world was adequately ready, Scott did what he had been told every startup founder should do: he hired a lawyer to make things official.

The list of legal tasks to handle was not long. Scott just needed to incorporate Mune as a company, bring on early employees, and protect IP via a design patent. But when the bill for this work arrived, Scott’s world—and the future of Mune—started to wobble. The fee was half of Mune’s bank account balance.

While legal fees threatened Mune, they sparked in Scott a new ambition. Only a couple of years later, Scott met co-founders Daniel Di Maria and Matt Mayers, and the three of them founded Spellbook to make business legal work more efficient and less tedious. It took many pivots before they finally found product-market fit (PMF), but now they certainly have:

• Spellbook raised a $50M Series B from Keith Rabois at Khosla Ventures in October 2025.

• Customers include commercial and in-house law teams at companies like eBay and Kennedys.

• The team moves “comically fast,” avoids meetings, and is hiring for 25+ roles.

The underlying goal for Spellbook has been the same since day one. “We want to make legal work 10x more efficient,” Scott said. And if the lawyers who use the product are to be believed, Scott & co. are doing a pretty good job at achieving that goal. Their product today is an AI tool that helps lawyers draft and review contracts much faster, so they can get more done and focus on higher-leverage work.

What is it like to work with the team at Spellbook? How effective is the product? Who is the right fit to join? I wanted the answers to these questions, and more, so I sat down with the team. Here is what I learned.

The lawyer

“Is this really what I went to law school for?”

It was a cold 11 P.M. in Toronto and Daniel Di Maria was still staring at a contract in Word, the tiny text now stinging his eyes. Fresh out of law school Daniel had joined a local firm, full of hope that working in law might be like what he dreamed about as a kid. Instead, he was copying and pasting legalese in a corporate contract for a company he would probably never think about again.

Daniel’s story is one that any commercial law professional would recognize: high hopes marred by boring contract work. Reviewing, revising, analyzing. Digging through thousands of words of text to find answers to important questions, to spot potential discrepancies, to mark potential risk factors.

Sadly for the lawyers involved, contracts are important; they are perhaps the most important part of commercial law. “$100T runs through contracts every year,” Scott told me. “They touch entrepreneurs, HR, sales, procurement, real estate agents, lawyers – everyone. They’re a bottleneck that prevents us from working together efficiently.” It would be nice if there were a way to speed all of this up.

When Daniel met co-founders Scott and Matt in 2016, the pieces snapped into place. Scott, stung by the Mune incident, was frustrated about how expensive legal work was. But Daniel knew first-hand that lawyers bill so many hours because the bureaucracy of law takes a lot of hours. It is not that lawyers would like to spend all of their productive hours reading minutiae in contracts; it is that they have to. The one way to fix both of these problems would be to make contract work significantly more efficient.

Contracts are the rational place for Spellbook to be building. Every company uses them, the market is huge, and they are ripe with low-hanging fruit to take care of; nobody wants to spend dozens of hours bogged down in drudgery and sentence-level detail. Spellbook’s AI, as it happens, is particularly great at handling this sort of work. Today’s AI models make right now a perfect time to be tackling this problem.

“With generative AI, the legal field finally has a tool that will empower lawyers to make their services more accessible to clients,” Matt, co-founder, said. “This industry touches everyone on the planet.”

The long road

The headline on Spellbook’s website today is straightforward:

Draft and review contracts 10x faster, just like magic.

(Given the company’s name, this is perhaps the one time we will excuse a “like magic” headline.)

But the path to landing on today’s version of Spellbook—now used by giants like eBay, Nestle, Crocs, Herzog, and more—was long and filled with pivots. The company was founded in 2017 under the name Rally, and most of their product revolved around document templating. This followed the same thesis that Spellbook still follows today (making legal work more efficient), but there was no AI in 2017. Templates were more rigid. Lawyers told Scott that Rally was a useful tool for junior lawyers, but that it did not capture all the nuance of high-level drafting. The company survived for years, but did not take off.

It was only when the first GPT models were launched that Scott, Daniel, and Matt got a new idea, sparked by seeing Github Copilot’s success in 2021: What if we made an AI copilot for commercial lawyers?

And so Spellbook was spun up as a side project for Rally. More than 100 landing pages and 200 growth experiments later, Spellbook took off in 2023. The waitlist surged past 30,000 lawyers in a matter of weeks, and it wasn’t long before massive commercial firms and in-house teams at Fortune 500 companies were coming on board. Scott, Daniel, and Matt had found product-market fit.

One of the core values at Spellbook is to work with a “beginner’s mind,” Matt and Scott both said. “It’s important we don’t get into an ‘expert’s mindset’ where we close our minds to too many possibilities.” This idea, which comes from the Zen buddhist concept shoshin, is perhaps the only reason that Spellbook exists today as a company. I have personally known founders who have given up after a couple years of slow growth; who have called it quits after a few successful pivots. And yet, 8 years later, here is Spellbook: a company that has worn many faces and, now, wears a winning one. Shoshin.

Today, more than 4,000 in-house and law firm teams use Spellbook to help them get contract work done faster. It was not particularly easy to get these people on board. “Lawyers are uniquely skeptical, detail-oriented, and protective of their time,” Matt told me.

To address this, Spellbook is very customer-obsessed. “The people who perform best at Spellbook are the ones who tie their work directly to productive quality and customer trust,” Aman Samra, the company’s VP of finance, said. In other industries, shipping a ‘good-enough’ product with obvious flaws and bugs might get you early traction. In law, it might kill your momentum entirely.

The StarCraft approach

It is after-hours, but the whole Spellbook team is still together. They have gathered for one purpose: to watch their CEO, Scott, play a game of StarCraft against a world professional.

One measure of a player’s skill in StarCraft is their actions-per-minute. Issuing a lot of commands (actions) per minute means you are likely a high-level player. Issuing too few could mean you are too slow, and will likely lose the game—because advantages compound quickly in games where you can reinvest your profits. In this company-wide demonstration, Scott is unfortunately on the losing side. His 110 actions-per-minute are crushed by his opponent’s 200. “Scott is genuinely good at StarCraft, but it was hilariously fun to watch him get crushed by a pro,” Daniel said.

After the game, Scott speaks briefly about the whole point of the exercise: we need to be a company that takes a lot of actions per minute. Companies that can do this will win. Companies that are too slow, who think too hard before doing anything, who wait too long to get things done—those companies will lose due to competitors out-compounding them.

The StarCraft approach seems to have been internalized through out the company. “Our culture is comically fast,” Chris Gardner (CSO) said. “That phrase is not a joke, it’s a real operating principle. We believe speed wins. While StarCraft’s APM is not literal at Spellbook, it’s a metaphor for moving through work with high output and momentum. Someone who can make a lot of good moves quickly will usually have more impact than someone who moves slower, and that impact compounds over time.”

Work at Spellbook tends to get done asynchronously, when possible, and meetings are kept to a minimum outside of the company’s mandatory all-hands. Constant live sync and check-ins are not common, and Scott, the CEO, “prefers written updates and likes to respond on his own time, which forces better thinking and reduces noise.” This approach usually only works when there is a deep level of trust between everyone on the team—which seems to be true, in Spellbook’s case.

The attitude toward mistakes, a normal side-effect of a high actions-per-minute mindset, is that “small mistakes are fine. They are recoverable. What matters is the average. If you are making many more right moves than wrong ones, you win in the long run.”

Should you join Spellbook?

There is an earnestness that comes through the bits and pieces I heard from the Spellbook team. This is a company where using a 1998 video game as a metaphor for how work should get done is taken seriously; a company that will list a Buddhist philosophy like shoshin as one of their most important values; a company that is not embarrassed to have fought for such a long time before reaching PMF.

People who struggle at Spellbook, Chief Strategy Officer Chris Gardner said, usually just need a different kind of environment. “One mismatch is someone who needs everything to be fully defined before they start. Another is someone who equates speed with recklessness. Another is someone who prefers mature operating systems over creating new ones.” Spellbook defaults to doing things, not thinking about doing them. If that does not resonate, the company may not be a fit.

When I asked Scott who should join, he built out a rather clear framework:

• Smart creatives: If you consider yourself a creative first, you’ll find yourself surrounded by a lot of like-minded people. We keep process to an absolute minimum.

• High autonomy: People who can take a lot of ownership and direct themselves.

• AI hackers: People who love running crazy, creative experiments with AI.

• Curious about legal: People who at least have a curiosity about the legal domain.

If you check all of those boxes, great. If you check even one, though, or a few, it may be worth checking out Spellbook’s open roles. They are currently hiring across sales, product, operations, marketing, finance, engineering and more. Most of these roles are remote. My bet is that you can find something. If you do apply, consider getting creative about how you do it, and consider sending a cold email. The more you can do to show the Spellbook team how you will actually be valuable to the company, the better.

Thank you to Spellbook for supporting Next Play and telling us what it’s like to work there.

TL;DR

If you only have a minute or two, here’s what you should know about Basis.

• Extracting value from AI means building useful tools. Basis believes that real AI productivity revolution isn’t magically unlocked by the existence of LLMs—it’s about smart, effective implementation. Basis is getting there by building AI agents for accounting that can work like teams of junior accountants 24/7.

• Accounting all the way down. Basis is building for accounting because they believe accounting underpins all financial (and much non-financial) decision-making. And because accounting is facing a crisis: the work is harder than ever, and there are fewer people than ever to do it.

• Real results, not vaporware. Basis already works with many of the top 100 accounting firms in the United States. Those firms are seeing serious reductions in manual work, monthly closes in 5 days instead of 15, and accountants focusing on strategy instead of manual tasks.

• A real AI-first company. Sometimes it is unclear what companies mean when they say they are “AI-first”. Basis has created a unique sort of internal team—called Atlas—with one single mandate: make each employee at the company 100X more productive.

• Intense, in-person, and high-responsibility. Basis works five days a week in their NYC office. The founders are first in, last out. The team frequently eats (non-obligatory) dinners together in the office. The culture is a fit for people who optimize for “raw impact” and who are comfortable with “enormous responsibility.” Basis raised a $34M Series A from Khosla Ventures in 2024, just one year after their seed round.

There is an ancient myth that our world is carried on the back of an enormous turtle. If you found yourself in conversation with someone who believed this, you would naturally ask: “What is underneath the turtle?”

And they would naturally reply: “It’s turtles all the way down.”

Talk to the team at Basis and they may tell you it is not turtles that are supporting our world, but accounting. They do not believe that to be an exaggeration. “One of the first things [the founders] Matt and Mitch sold me on is the idea that accounting is the bedrock of our economy,” Luke Bradley (a generalist at Basis) told me. “Accounting is how we compress real-world complexity in financial data, which underpins all decision-making. If LLMs progress unlocks accounting in the way the early Basis team expects, it could lead to an enormous improvement in the overall quality of decision-making.”

The world, or at least the financial decision-making part of it, is accounting all the way down.

Every company understands this, which is why internal accounting departments are so important and why accounting services generated $145B in revenue in the U.S. last year. But accounting is not always an easy job, and common complaints have gone mostly unfixed for years:

• Globalization, the internet, tax codes, and regulations have made the job more complex.

• There can be long hours and insane crunch for month-end closes and during tax season.

• There is a ridiculous amount of tedious, manual data entry, calculation, and fact-checking to do.

• Accountants are mostly unsung heroes (nobody says ‘Wow, they’re an accountant, that’s so cool*.)

(*Except, maybe, the team at Basis.)

As accounting has simultaneously become more complex and more important over the past decades, interest in the profession has waned. Fewer people than ever are taking CPA exams. Senior accountants are retiring with nobody to replace them. Companies and firms are outsourcing to other countries (like India) when they can’t find domestic talent. Quality, in some cases, is decreasing.

To solve this problem, founders Matt Harpe and Mitch Troyanovsky started Basis to build AI agents for accounting. If accounting is at the center of decision-making, then improving accounting can unlock a substantially better world. So far, they seem to be making some inroads. Basis:

• Raised a $34M Series A from Khosla Ventures back in 2024.

• Works with many of the top 100 accounting firms in the U.S.

• Cares a lot about being AI-native. They have an internal team, called Atlas, with just one mandate: make each employee at the company 100X more productive.

• Has built an impressive team with an intense, in-person, high-responsibility culture.

​

I wanted to learn more about what Basis is building, why, and what it might be like to work there. So I chatted with the team and will do my best to present my learnings below. If all you want to know is whether or not you should join, you can skip to that part here. But I’d recommend reading the rest first: Basis is approaching company building in a way that is worth learning about.

Building useful tools—not just more software—for accountants

What if every accountant had a team of 100 junior accountants that could work 24/7?

The answer is obvious: they could be insanely productive. Accounting is experiencing some real tension (more complexity and fewer people to handle it!), but an innovation like this could completely transform the industry. It could make each individual accountant much more effective. It could improve the quality and speed of accounting overall, which could improve the quality and speed of all decision-making.

So, in 2023, Matthew Harpe and Mitchell Troyanovsky went about making this happen. Their $3.6M seed round was led by Better Tomorrow Ventures, and their $34M Series A—just a year later!—was led by Khosla Ventures with participation from Keith Rabois, Nat Friedman, and others.

The goal at Basis is not to give accountants more software. “Every accountant already has dozens of software tools,” the Basis team writes. What accountants really need is useful tools. “Tools that can understand accounting, learn specific client needs, work independently, and take real actions.” This approach is working, and many of the top 100 accounting firms in the United States are already seeing results with Basis.

All of this information is exciting. But one of the more interesting things I learned when talking to Basis is that they take AI very seriously; the team is not just building AI agents for accounting, but for themselves.

Fighting the consensus to build a truly AI-first company

Plenty of companies today claim to be AI-first. But when you dig deeper, that tends to be a hollow phrase: sure, some employees use ChatGPT and the engineers use Cursor, but these days that is the consensus. There is nothing particularly impressive anymore about having AI involved in your workflows.

Basis is one of those rare companies that has thought deeply (from “first principles,” they told me) about what it means to build a company for the “AGI era” and has made some non-consensus choices.

Perhaps the most unique decision? There is an entire internal team at Basis, called Atlas, who have one mandate: make every employee at the company 100x more productive.

Matthew Busel, the Product Lead for Atlas, describes the team as follows: “AI is already incredibly smart and only getting smarter on the road to AGI. So why can’t it do much of any work within a company? The reason is that it doesn’t know anything about your business - aka it doesn’t have the right context. The Atlas team is solving that problem to make Basis the most AI-native company in the world where hundreds of agents can work alongside our team as we try to scale as fast as any company ever has.

Building an “internal agent team” (another way to describe Atlas) is not the standard now, but according to Basis, it will increasingly become so as we approach AGI. In the PC era, someone had to create the first IT team. In the internet era, someone had to create the first data team. Atlas, if it is a successful bet, could be one of the first internal agent teams in a world that will eventually be full of them.

“The fact that my team even exists,” Matthew said, “is a great example of the uniqueness of this company. It’s one thing to tweet about how you think an AI-first company should operate. It’s another to actually do something about it. Matt and Mitch are funding a team to build a future AI-native organization way earlier than you’d see at any other companies.”

Matthew is correct: it is strange to see a Series A startup investing deeply in an internal AI agent team; in people who are not building the customer-facing, but building products for people who build the product. If the bet pays off, Basis will in 20 years be the sort of company people write about in essays on the ways companies changed during the AI era.

And if Basis achieves even a fraction of that “100X more productive” number internally and for their customers, it would be hard to understate the impact they might have on accounting and on the world.

A generational mission comes with generational responsibility

The Basis team works five days a week in their NYC office. They describe working there as carrying “enormous responsibility.” The company pays for (non-mandatory) dinners at the office for everyone from Monday to Thursday. Matt and Mitch, the founders, are “the first ones in and the last ones out every single day.” There is “no such thing” as a normal day at Basis.

This is a startup that sees an enormous opportunity over the next 18-48 months and enjoys working with ambitious people who take that opportunity seriously. People who recognize the responsibility that comes with that sort of mission. There is endless theorycrafting we could do here, but the best thing to do is hear straight from the people who actually work at Basis. So I asked around:

• “The team is intense but full of positivity. Everyone here is impressively smart and excited to build something special, but highly collaborative and encouraging. It’s genuinely fun.”

• “One of my favorite meetings at the company is ‘Wednesdays with Mitch’, where once per month our co-founder Mitch takes questions from the whole team and has a transparent, open discussion about Basis, our strategy, and how we think about things.”

• “There is an emphasis on minimal meetings. Many ad-hoc conversations around the office. Culture of (respectful!) interruptions for unblocking yourself or others.”

• “The team is fun. I enjoy lunch and dinner with them in the office — cracking jokes, making plans for after work.”

• “We might spend one morning architecting a new system for agents. By the middle of the day we’re sitting side-by-side with our deployed intelligence team to deeply understand a workflow we think we can improve with AI. In the afternoon we’re building an agent that solves the problem deployed intelligence shared with us. And the next morning we’re shipping it live for feedback.”

Talking with Basis, the impression I got was that the people who work there work hard, sometimes work long (dinners in the office imply longer-than-usual time at work!), and have a hell of a lot of fun doing it.

Should you join Basis?

There are truisms that most every startup feels that they have to say when they talk about the people they are looking for. Hard-working, high-agency, low ego, collaborative. You would be hard-pressed to find many startup job descriptions today that do not contain at least some of those words. But phrases like very intense and enormous responsibility are a bit more rare; they are intentionally dramatic.

I have not worked at Basis and so I cannot say with certainty whether those phrases are true, but I can say that using them signals something about Basis. This is, according to co-founder Matt, “a small group of ambitious people who believe there’s a unique opportunity at this moment in time with AI. We want to build great things, which requires intensity but is also really fun.”

People who are good fits at Basis “do not want a typical 9-to-5” and “do not care about status or title, but instead raw impact.” Working at Basis means buying into what they are building and truly caring, not “viewing work as a flywheel for life.”

When you apply, you can expect “a fairly unique interview process,” I was told. “We do our best to do away with the interviewer/interviewee power dynamic and instead try to collaborate directly on problems and see what it would be like to work together.” Someone else told me that interviews “put an emphasis on solving real problems faced in the office.” Interviewing with Basis is “comfortable,” someone else said, and feels “more like showing up at work and not like you are in an interview.”

So, if you would like to join a (truly) AI-first company transforming accounting, how should you apply?

You could start by looking at the Careers page and applying through traditional channels: Basis is currently hiring across product, marketing, accounting (of course!) and more. In addition to a formal application, you could send a short, useful email to the person hiring for your role explaining the value you think you can add. And if you want to try extra hard or get creative, you could stand out by trying ideas from a list like this.

Thanks to Basis for interviewing with us and for supporting Next Play.

If you only have a minute or two, here’s what to know about Chalk:

• The data platform for inference. Chalk powers real-time AI/ML models for companies like Whatnot, Iwoca, and Mission Lane with drop-in infrastructure for data teams. Before Chalk, sophisticated ML systems were primarily built by internal teams in the fintech industry, spurred by compliance, KYC, and fraud detection needs.

• Three repeat founders with deep fintech expertise. CEO and Co-founder Marc Freed-Finnegan launched Google Wallet and sold his company, Index, to Stripe. Co-founders Andy Moreland and Elliot Marx spent time at Palantir and Affirm, respectively, then built Haven and sold it to Credit Karma. The three identified an opportunity to systematize the AI and ML infrastructure pioneered by the fintech industry by creating an ultra-fast, scalable, and industry-agnostic data platform.

• Absurd momentum in 18 months. In May this year, they raised a $50M Series A at a $500M valuation from Felicis. Chalk doubled its team in the last year, from 30 to 60, and plans to double again in 2026 with open roles across design, engineering, go-to-market, operations, and more.

• The best in the world. Everyone at Chalk is obsessed with moving fast, 5 days a week in-office. People who are a good fit at Chalk are “truly aspiring to be the best in the world at what they do,” according to their hiring ethos. This, naturally, means caring an unusual amount about your craft(s) both in the office and outside of it. “We’re all a similar type of nerd,” one of their engineers said. Equally driven is Chalk’s industry-leading go-to-market team, led by VP of Revenue Alexandra Kane, which has accrued some incredible enterprise logos.

We often write about how you might go about getting hired at a startup. There is a follow-up question here, though, that I sometimes get asked: How do I identify a startup that is actually good?

We all know the stats. Most startups fail. There is risk! More than 15,000 startups were funded in 2024, and yet only about 20 companies per year end up eventually exiting for at least $1B. That puts your odds at about 1 in 750 to pick a company that is really going to matter.

Over the course of my conversations with the team, I have recently begun to wonder whether Chalk—a data platform that powers AI/ML applications—may be one of those companies.

On paper, Chalk is undeniably impressive. They’ve achieved product-market fit, with dozens of enterprise customer testimonials. They recently raised a $50 million Series A at a $500M valuation, led by Felicis (with participation from General Catalyst, who also co-led their seed round). And they are hiring for roles in SF in every department: engineering, marketing, sales, operations, and more.

The culture, according to their founders and team, is relentless and high-velocity. “We’re finishing entire pilots before our competitors even complete their security reviews,” Andy told me.

In this Next Play Spotlight, I try to learn more about Chalk. What do they do? Why have they grown so quickly? What is it like to work there, and should you consider joining? All that and more below.

AI is booming, but it needs a faster data layer

It is somewhat obvious to state that success looks kindly on companies that (successfully) build core infrastructure that is upstream of a lot of value in the world. AWS. Stripe. NVIDIA. In many ways, Chalk is angling to be that sort of foundational company—the kind that powers trillions of dollars of useful tools.

So what does Chalk actually do? What is the market?

Well, if you are a software engineer or work in AI, “the data platform for AI + ML,” as Chalk describes themselves on their homepage, may make immediate sense to you. However, if you are not in this space and that line does not make it clear to you what their product does, here is a short explainer.

Imagine you work on transaction fraud and credit underwriting at Ramp. Your principal goal is to make effective decisions with data: every time a transaction occurs you want to make sure it’s legitimate, and every time someone applies for credit you want to make the right decision about how to proceed. You use machine learning models to do this. These models need data—lots of it—in order to work.

The problem is that the data to power your fraud and underwriting models lives in two dozen different places; separate databases, live signals, API calls. Your model needs all of this data to make a decision, and it needs it in (literally) 10 milliseconds so that the customer has a good experience.

Chalk is the data platform that instantly connects this data to your machine learning models, unlocking what’s known as “real-time inference.” In plain terms, Chalk’s data platform is what allows models to make smart decisions (inference) in real-time based on data.

This is, as you might expect, quite a large market. And it’s growing fast. “Most of the friction in AI,” wrote Aydin Senkut of Felicis, who is now on Chalk’s board, “comes after training. Real-time inference is complex.” As AI/ML continues to grow, so will Chalk’s TAM.

And the market is not AI hype. It’s solid. Chalk’s success does not rest on the kinds of things that AI skeptics often bring up (like whether Open AI is going to have the impact their valuation suggests). Instead, Chalk powers real, often straightforward tools used by real, profitable companies to add value.

• Whatnot uses Chalk to power real-time recommendations (computing >300M features / second).

• Mission Lane uses Chalk for real-time credit approvals and fraud detection.

• Apartment List uses Chalk to run its dynamic search recommendations.

• MoneyLion uses Chalk to unify ML development, accelerate feature delivery, and reduce time-to-production.

Chalk benefits massively from the rise in AI/ML models, many of which need real-time inference (and thus Chalk) in order to be useful. But, it is not hamstrung by AI speculation or an AI bubble; Chalk’s business does not hinge on AGI being achieved, or whether chatbots are truly ‘the next internet.’ That’s refreshing.

But, before I get ahead of myself—why was Chalk founded in the first place?

The founding team behind Chalk

Perhaps the best way to describe Marc Freed-Finnegan, Elliot Marx, and Andrew Moreland (Chalk’s co-founders) is to say that they are, and have always been, fantastic problem-solvers.

Marc spent many years at Google, where he helped to launch the first version of Google Wallet. He went on to start Index, which Stripe acquired as its in-store payment solution—now called Stripe Terminal.

Elliot and Andy, meanwhile, met at Stanford. Elliot began his career at Affirm, where he built the early risk and credit data infrastructure (part of the inspiration for Chalk). And worked at Palantir on large government data infrastructure projects. In 2017, Elliot and Andy joined forces to found a company called Haven Money, which Credit Karma later acquired to power its banking products.

If you asked me to write down a perfect-world version of what kind of people I’d want to start a company like Chalk, I would probably tell you I’d want people who (1) have all had successful exits and (2) have experience building quality data infrastructure. In Chalk’s case, that dream is reality.

And so far, things seem to be working. Since raising their $10M seed round in December 2023, Chalk:

• Built a product that has won “every single head-to-head pilot” against well-funded competitors.

• Signed big customers like Whatnot, Sunrun, Socure, MoneyLion, and Apartment List.

• Raised a $50M Series A at a $500M valuation (in May 2025).

• Grew the team to 60 people.

All the while, they’ve scaled their ambition, problem-solving, and velocity with a company culture that—at least from what I can tell as an outsider—seems quite effective.

A low-ego culture driven by velocity

Any self-respecting startup will tell you, at least in broad terms, that they like to work hard. That they like to get shit done. That it’s important to be a self-starter and work fast. That’s kind of the startup bible. Often, though, much of this falls apart upon closer inspection. Chatting with the Chalk team, I got the impression that everyone is dead serious about their values, particularly their problem-solving and velocity—and that the company would not exist without them.

There were a few themes that showed up when I talked to the Chalk team; below I’ll organize them, with quotes, so you can get a strong sense of what this company is all about.

1: Chalk is unusually obsessed with their customers. “I read almost every single customer message we’ve ever received,” Andy told me. “ Every morning, when I wake up, the first thing I think about is our customers. By the time [Databricks execs] finish their first meeting, I’ve gone through Slack messages from 30 customers.”

2: The whole company is driven by velocity. “We’re collaborative and low-ego,” Sai Atmakuri, a software engineer, told me, “which means we are able to plan and move really quickly. When a project that’s been in the backlog is suddenly needed to get a pilot over the line, or a customer encounters an issue, we figure out how to build that out ASAP.”

And according to Andy, Chalk is so fast that they “finish entire pilots before our competitors even complete their security reviews. Chalk customers ship to production faster because we refuse to be the rate-limiting step in their process.”

At Chalk, you’re expected to ship no matter what. “One of our core values is incrementalism,” Elliot said. “Just do something that makes forward progress and ship it, even if it doesn’t do everything or even if you’re not going to use it yet. You’ve got to get the door out consistently.”

Elliot recognized that incrementalism has “a dirty name, not just in engineering. But everything should be incremental. Our whole world should be incremental. Take a small step in a direction, then observe and see if you like it or don’t like it. You need to be honest with yourself about when you’re wrong and when you’re right.”

3: Everyone at Chalk is a self-starter. Caroline Parker, a sales development representative, told us that, “You need to be the type of person who’s hungry and willing to learn, do a ton of reading, and immerse yourself in material. You need to be willing to dive right in. If you have an extra five minutes, you might ask ‘How can I help?’ ‘Where can I be useful?’. You should always have something to do, and you shouldn’t always have to ask. You should just pick it up.”

This was echoed by everyone we talked to. Melanie Chen, a forward-deployed engineer, said that when problems come up, “we wear many hats and just solve them. That’s a Core chalk trait. You can’t expect everyone else to solve your problems for you.”

Brian Sussman, operations manager, agreed. “You definitely get out what you put in. The people who succeed here give it their all. It’s easy to get a job [anywhere], but this is the place to build a career.”

4: The culture at Chalk is all-round obsessive and a bit nerdy. “We’re all a similar type of nerd,” Sai said. “There’s folks who are into high-performance computing, chess, poker, board games. There’s a good overlap between people who are interested in geeking out about all these niche things, and that makes for a very fun and friendly environment.”

“I bought a Nintendo Switch for the office,” Andy said. “Turns out one of our engineers ranked among the top competitive Smash players in SF and Pittsburgh. A great thing about Chalk is how people here are among the best in the country at something totally random. It’s great to see people choosing to stay until 9pm for Smash Bros or Mario Kart tournaments. That’s what happens when you make the office somewhere people want to be.”

Should you consider joining? And if so, how?

If you have made it this far in the essay, I’d wager you have a decent idea about whether Chalk is your kind of company (intense, obsessive, nerdy, high-velocity, effective). But there’s some more information you should be aware of that, if you are the right kind of person, should get you even more excited.

One feature of the company is that they are in the office 5 days a week. Ten years ago this was the default, but now it feels like something of a contrarian opinion. “We’re [in-person],” Andy said, “because we want to iterate really, really fast. Remote doesn’t work for high-performance teams,” he told us, because decisions can’t get made fast enough, collaboration is more difficult, and because people feel more accountable when they are working on projects in-person.

You may also end up working longer than the traditional 9-to-5. “The people who write software here, they’re not just writing it 9 to 5,” Marc said. “They’re going home and they’re still writing software. And maybe it’s for Chalk, or maybe it’s for a side project, or maybe it’s for a competition they’re part of. The point is that they really love what they do. They’re aspiring to become one of the best engineers in the world.”

So that much is clear—if you want to fit in at Chalk, you should be almost unreasonably (from an outsiders’ perspective) dedicated to your craft. You should be ready to work hard, but more than that, you should want to work hard.

And if you are still in doubt, I think Marc summarized it well:

“We’re looking for three things. Number one, we are looking for people who are really, really smart. People who can figure out new things because the world around us is changing and we have to figure out things all the time. Number two, we are looking for people who can communicate really well. That makes writing great software, because that’s our language. But even if you think about other roles at Chalk, communication is essential. Number three, we are looking for people who are aspiring to be the best at what they do.”

If you have read all of this and think that Chalk sounds like the right place to build your career at, check out their careers page. Of course, you don’t have to stop at sending in a generic application. You can also directly email the person who’s hiring for your role. You could record a Loom video about yourself, which is what Brian did to get hired. You could directly DM somebody on LinkedIn. You could build something that might be valuable (or useful, or interesting) and send it to Chalk. You could talk to their customers and bring them learnings. Or you could try one of the ideas on this list.

If you only have a minute or two, here’s what you should know about Valence and their AI coach Nadia.

• Quality, 1:1 coaching at scale + a new platform for HR. Every enterprise wants their team to be more effective, but traditional coaching is expensive, limited, and impossible to scale to every single employee. Valence built their AI coach Nadia to give personalized guidance and world-class coaching to every single person (from frontline employee to C-suite exec).

• A coach is just the start. When every employee has an AI coach, it becomes a new platform for HR to deliver learning programs that accelerate careers, make performance reviews a tool that actually improves how we work together, and to help companies and teams stay aligned in times of change and uncertainty.

• One of the most advanced AI systems in use at global scale at the world’s largest companies. Nadia was the first AI coach to market and is the leading AI coach today in the Fortune 500. Valence’s earliest customers, including Experian, Prudential, and General Mills, are almost three years into their coaching journey and are using coaching at global scale not just in pilots.

• A team that knows exactly how to work with the world’s biggest companies. Co-founders Parker Mitchell, Alex McMurray, and Levi Goertz both worked at McKinsey, and have deep experience in enterprise consulting. Former product and engineering leaders from Slack are designing and building Nadia.

• A fresh $50M raise and strong momentum. Valence recently closed a $50M Series B led by Bessemer, and they’re hiring across New York, London, and Toronto. The culture is collaborative and ambitious, and employees who succeed take ownership and drive real world outcomes (“not for people who need structure,” I was told).

There is good reason for the hype that early founding teams get: it is hard to hide being ineffective when it’s you and five other people on a team. The founder(s) can set the standards and verify them manually. Every piece of work is visible to everyone. Taste and productivity become automatic; the whole team can work as one productive unit. This is good, for a while. But then the company scales.

As a company hires 100, 1,000, or 10,000 employees, no single person can have visibility into everything. Some wings of the organization are more productive than others. Culture morphs and becomes asymmetric across people, teams, functions. This quality slippage, almost an expected outcome for enterprises, is frequently cited as the reason large companies become slower.

What would it take to make the biggest companies as agile and productive as the most ambitious startups? Well, if you were going to design a solution without any limitations, one idea would be to hire one full-time coach for every single employee at your company. This coach would:

• Be extremely smart

• Read every single coaching book and know every framework

• Have full context on everyone at the company

• Be available 24/7

• Think and produce ideas way faster than any human

• Maintain company culture, standards, and priorities across every single employee

Of course, this is an impossible idea. No company can afford to immediately double their headcount, and no coach can hold infinite context in their heads (nor can any coach be available 24/7). As it stands, traditional wisdom goes, companies cannot execute on this ambitious but impossible idea.

… Or can they?

Valence, co-founded by Parker Mitchell, Alex McMurray, and Levi Goertz, has built an AI coach, Nadia, that can truly understand and anticipate the needs of employeeso make everyone—especially managers and leaders—more effective and collaborative.

So far, so good. Valence:

• Is the leading AI coach in the Fortune 500 with customers like Delta, General Mills, and Prudential.

• Raised a $50M Series B led by Bessemer Venture Partners in August 2025.

• Built a team of deeply ambitious, mission-driven first principles problem solvers. Their product and engineering leadership came from Slack, the last software to reinvent enterprise work.

I wanted to dig deeper into what Valence is building, why, and who might be a good fit to work there. So I sat down with the team to figure out why they’re building what they are—and whether you should join.

Thanks to Valence for chatting with me and for supporting Next Play.

Coaching is effective, but expensive and difficult to scale

It is not a novel practice to hire a coach, especially for your executives: today, the market for doing so is about $103B in the United States. Companies know that good coaching can make their leaders, and thus their companies, more effective. So they shell out a lot of cash to coaches.

Traditionally, though, coaching has been rather limited. Perhaps you hire some expensive outside help to coach a handful of your VPs who you think could benefit the most. Perhaps you build an internal coaching program for high-performance employees, like what Prudential (now a Valence customer) did. Whatever you do, though, you typically aren’t doing it across the entire organization for every manager and for every employee; at least not in any personalized capacity. Good coaches are hard to find, have limited availability, and are expensive. So you do the best with what you have—targeted 1:1s for certain execs, group workshops for some, and nothing for the rest.

But what if every single employee had a world-class coach on demand?

This is the version of the world that LLMs could make possible. On paper, a technology like ChatGPT is a perfect solution to the coaching problem—it is basically an infinite text generation tool that can sometimes come up with really smart answers and can hold a lot of context in its head. (And many people already use ChatGPT on their own for coaching-like needs.)

If ChatGPT is so miraculous, though, then where are all of these 10X companies? Why doesn’t every employee at every company already have this genius, 1:1 coach to help them?

This is where reality gets complicated and where Valence may have an edge.

Back in August, MIT came out with a landmark report that 95% of AI pilots are failing. This lined up with what I’ve seen more anecdotally; big, hype-y promises about the potential for AI were clashing with the actual reality of what these software products could do. At many companies, this reality was disappointing. No, AI can’t (usually) replace all of your smartest employees 1:1. No, AI can’t magically do many of the things flashy marketing statements say it can. No, you are (probably) not going to build a multi-billion dollar company today by firing all of your employees and replacing them with virtual AI agents.

“The future is not humans versus machines,” Das Rush, Valence’s Head of Marketing, told me. “The future is humans with AI.” It is becoming increasingly clear that AI is not a silver bullet to use however you please, but rather a potentially useful technology that can be made actually useful when it’s developed by a team of effective people for a specific purpose (e.g. Cursor for coding).

“AI is ushering in the era of mass personalization,” Parker Mitchell, the CEO of Valence, said. And so Valence’s goal is both simple and timely: not to build a generic LLM tool, but to create truly useful and personal AI coach that can make every single employee more effective.

Building a personalized AI coach with a 98% pilot success rate

Valence has always been about making teams more effective and believes they are the essence of how modern knowledge work gets done. “Before ChatGPT came along,” Parker said, “our very first product was a paper prototype about understanding your team. I brought a dozen copies to a conference, sidling up to attendees, and [sent] them a photo of their answers along with some advice.”

ChatGPT launched in 2022, shortly after Valence raised a $25M Series A. They quickly pivoted to building personalized AI coaching for enterprises and by the end of 2022 their coach was in its first enterprise pilots. Today, their flagship product Nadia is used by 50+ of Fortune 500 companies—is an AI coach called Nadia.

Nadia’s secret sauce isn’t a model, it’s context and memory. Out of the box, Nadia is purpose-built to draw from the best coaching practices and frameworks. Valence then specializes in customizing Nadia to each company, training her to coach to the specific goals, culture and dynamics of the company. But most powerfully, Nadia is designed to build more context with each coaching conversation, so that the more she coaches, the better she gets.

Nadia does what a good human coach would do. She looks at an employee’s tendencies. At what they have going on. At what their relationships are with other people at the company. At what they’ve struggled with (and done well at) in the past. Then she provides effective, objective analysis and feedback to help the person become more effective (based, of course, on the company’s priorities).

Nadia is working: 98% of Valence’s pilots are successful. Prudential 9X’d the reach of their global coaching program with Nadia. Experian had a pilot fully up and running within 6 weeks, and has now expanded access to Nadia to a third of their workforce. Delta used Nadia to improve 97% of their yearly performance reviews. Real enterprises with real problems are, seemingly, seeing real results.

These numbers are substantial outliers—remember, 95% of AI pilots seem to be failing. Valence’s traction so far appears not to be the norm, but perhaps something rather exceptional. And, coupled with their $50M Series B just a few months ago, they are in a position to hit the gas pedal.

But how are they doing it? Why is Valence selling to Fortune 500 customers and succeeding with almost all of their pilots when the industry standard appears to be far worse? The answer may lie with the impressive team and hard-working culture they have built.

A sharp, mission-focused team that’s scaled companies before

Perhaps the reason Valence has been so good at selling to Fortune 500 companies (often a difficult leap for startups to make) is because they practice what they preach. It’s a culture deeply committed to being the most AI-first of workplaces and using AI to reinvent how they scale as a company. Co-founders, Parker Mitchell, Alex McMurray, and Levi Goertz are obsessed with rapid-fire experimentation and solving problems from first principles. If you need to “operate from a list of best practices,” I was told, Valence is not your place. (Refreshing!)

The team is high-caliber, too. Ana Martinez, Valence’s Engineering lead, was formerly the Director of Engineering at Slack. Das Rush, who is the VP of Marketing, formerly ran content for a16z’s growth fund. Other teammates (like Kira Luscher, growth, Mehwish Panjwani, client solutions, and Ellie Wildman, strategy) have deep backgrounds in consulting and industry, but a hunger to apply abstract theory to real world deployments.

According to Daniel Owens (client solutions), “A lot of us come from consulting backgrounds, where the work is often theoretical. At Valence, there’s a refreshing shift—it’s all about real-life behavior change and measurable outcomes. You’ll see people spinning up an MVP in a day, running experiments with real users, and using what we learn to build better solutions immediately.”

This theme (collaborative, hard-working, focused) was echoed by everyone I talked to:

• “There is always a lot to do, but I have so much fun every single day. I’m learning a ton, sharing a lot of laughter with people who work here, and I’m genuinely excited to see everyone daily.” — Zainab Jangda, CS Associate

• “Every single employee at Valence is the person who carried the group project in school, while also being someone you genuinely want to hang out with. We are a team of reliably hard workers who i) are eager to grow and learn from one another, ii) will explore the most cutting edge tools to do so, and iii) want to build a generational product that people love.” — Kira Luscher, Director of Growth

• “Valence is highly collaborative and everyone is super nice.” — Ana Martinez, VP of Engineering

• “Our ability to work cross functionally, and view ourselves as one Valence team (versus siloed teams) is exceptional. Everyone is operating under tight deadlines, but the mentality is us versus the problem, which creates an expectation and behavior of problem-solving versus blaming.” — Ellie Wildman, Director of Strategic Accounts

• “Valence is intense, mission-driven, and deeply collaborative. It is a place for people who love solving complex problems at speed — where ambition meets humility, and everyone’s committed to building something that truly changes how people work. It’s a high-performance culture, but one grounded in curiosity, care, and a shared belief that growth is a team sport.” — Das Rush, VP of Marketing

These are generally good things to hear if you are interested in joining a startup that gets things done and cares about building a useful product. But I wanted to dig a bit deeper on who, exactly, Valence is the right fit for—so you can make up your mind about whether this may be a company worth joining.

Should you join Valence?

Over time, companies often start to develop a sort of “vibe”, a certain way of doing things that makes some people a better fit than others. I wanted to learn what kinds of people, exactly, would do well at Valence. Below are a few pieces of information I think you may find useful.

On routine: “There is honestly no average day [at Valence]. The job varies so much from one day to the next.” — Mehwish Panjwani, Director of Client Solutions

On meetings: “High signal, no fluff. People are really comfortable with each other, so we move quickly through agenda items, give direct feedback, and leave with a very clear list of next steps. There’s not a lot of posturing or over-formality—everyone is focused on making decisions or unblocking something.” — Daniel Owens, Director of Client Solutions

On what kinds of people would be a great fit: “Someone that is comfortable operating in ambiguity, and who wants to run towards problems, solve them quickly and iterate. Someone with a strong sense of ownership, who sets a fast pace for themselves. Someone who is scrappy, flexible, and collaborative. Someone who can roll with the punches. Someone who can operate with earnesty and urgency.” — Multiple people on the Valence team

On who would NOT be a good fit: “Anyone focused on ‘strategy’ without willing to get dirty in the execution. Anyone not using AI tools on a regular basis, or getting excited to experiment with new ones.” — Kira Luscher, Director of Growth

If you like “operating from a checklist of best practices,” or if you “need a lot of structure,” I was told, Valence may not be a great fit. But if you are ready to work hard, focus on execution, and collaborate across every team in the company—it may be worth learning more.

Valence is ramping up faster than ever after raising their Series B and is currently hiring for a number of roles in New York, London, and Toronto. You can apply via the link on their website, but if you want to try extra hard, you could also try one of the ideas here. Or you could (in addition to an application) simply email the person who’s hiring for your role explaining how you think you might be valuable.

TL;DR

If you only have a minute or two, here’s what you should know about Zingage.

• Healthcare where people actually want it. The vast majority of people want to age at home, not in nursing facilities or in hospital rooms. Zingage is automating home care logistics with AI so home care agencies can focus on delivering great care instead of drowning in scheduling, payroll, and documentation. It’s a straightforward mission: help people get better care in their own homes.

• Massive market, perfect timing. The home healthcare market is already worth $400B+ and it’s about to explode. By 2030, there will be 80 million Americans over 65—double what it was in 2010. And 93% of seniors have a chronic disease. Home care agencies are scrambling to keep up, but they’re hamstrung by clunky software (or no software at all). Zingage is solving that problem.

• Two products with real traction. Zingage built Operator to handle the back-office chaos (scheduling, calls, billing, documentation) and Perform to help agencies retain great caregivers. The results speak for themselves: Zingage has scaled to millions in ARR with just 15 people on the team. They already have 400 home care agencies using their products, representing 50,000 caregivers. They just closed a $12.5M seed round led by Bessemer.

• Stacked team of builders. Victor (CEO) previously scaled a real estate software company to $50M+ in volume. Daniel (co-founder) was the fifth growth engineer at Ramp. One of the founding engineers co-founded Bearworks and spent years at Datadog. The design lead was a founding designer at Tennr. Everyone on the team has done real, impactful work.

• “Be a pirate with responsibility.” The culture at Zingage is intense and high-agency. People here move fast, take ownership, and solve problems without waiting for permission. It’s not for everyone—it’s overwhelming sometimes—but if you want to learn how to be a founder or make a real impact on millions of people, this is the place. They’re hiring 10 engineers now, and to apply, co-founder Daniel wants you to send him a Loom with the most impressive thing you’ve built.

Does your startup make the world a better place?

For many companies, it is a harder question to answer than you might think. I have friends working at startups who, to answer this, would have to build a 7-step flowchart to show me how the startup has a positive impact on real people in the real world. “Well, we built an integration for A company, which helps B company, which in theory could make the world a better place by increasing the efficiency of C…”

Hmmm. Interesting.

With Zingage, answering this question is quite straightforward. There is only one step. Founders Victor Hunt and Daniel Tian and their team are building an agentic AI layer for home care agencies so that healthcare can improve for people at home who need it most. Now that feels like a mission almost anyone—even the most jaded tech startup skeptic—could get behind.

Of course, making the world a better place is not the only qualification you need to be a successful startup worth working at. You also need a product that can make a lot of money. Turns out that, so far, Zingage is doing quite well on that front too: they recently launched a $12.5M seed round led by Bessemer. This fundraise was built on the back of some impressive metrics:

• They already have 400 home care agencies (employing 50,000 caregivers) as customers.

• They have two functional, useful products, Operator and Perform, which we’ll cover later.

• The Zingage team is unusually high-caliber—filled with former founders and people who made serious impacts at Ramp, Uber, Datadog, Citadel, and Tennr.

But how did they do all of this? What are they solving? What’s the ceiling? What even is home care?

This essay will answer those questions, and more.

Healthcare is a huge market, and it’s coming back home

The term “home care agency” does not tend to come up in daily conversation. It is worth explaining the (very important!) industry Zingage is actually in before diving into their product and traction.

So imagine this: you just celebrated your 81st birthday. Life has been long, and good, but it has treated you poorly in recent years. You can’t walk like you used to; you fall often. The list of chronic disorders and general health problems you’ve been diagnosed with seems to get longer every week. You can’t do the things you used to do, and you need basic help (like dressing, walking, and taking medications) just to get through the day. It’s not the life you pictured, but then again, life never is. You wish your kids could help, but two of them live hundreds of miles away and the third is a single mother who works full-time. You are running out of options, and life somehow gets more complicated with each day.

You know you can’t handle things on your own or with your spouse, but you also don’t want to leave home. Spending the final years of your life in a sterile, unfamiliar nursing home or in hospital rooms anytime something bad happens does not feel like a satisfying way to end things.

What you, and the vast majority of Americans want, is to age in place. In a home you know and are comfortable with; a place you can continue to get the most out of life, to the best of your ability.

The situation I have just described is one you may end up in one day. It is also a situation you more than likely have witnessed with people in your life (parents, grandparents, friends, other relatives). These are the kinds of people that home care agencies exist to serve.

The $400B+ global home healthcare market is effectively built on two facts:

• Most people develop health problems as they age.

• Most people (77% of American adults) would prefer to age in place.

Hundreds of thousands of home care agencies—from small family-run businesses to large companies—exist to help solve this problem. And the market is growing at an insane pace:

• In 2010, there were ~40M Americans over 65. In 2030, there will be ~80M.

• Disease is on the rise, and today, 93% of Americans over 65 have a chronic disease.

And so the basic situation these days is that there is a massive home care market which will, more than likely, at least double (perhaps far more) over the next decade or two. In any industry, this would be a sizable, and challenging, opportunity. But in home care it is uniquely complex.

The reality today (not in 10 years, not when the market doubles, but right now) is that running a home care business is ridiculously difficult. These are not moonshot software startups with millions of VC dollars and years of runway. These are businesses with tight margins that need to be profitable and provide a high standard of care. And most home care agency owners get so bogged down with their back office work that they have little time to focus on growing their business or improving their quality of care.

Victor Hunt, co-founder and CEO of Zingage, knows all about this: he started Zingage because he saw his grandma struggle to run her own home care agency. “She spent more energy buried in scheduling and operations than on delivering care to the people she built the business for,” he told us.

Talk to any home care agency owner and they would tell you what they wish they could do: focus on delivering the best possible care to their clients and patients. Then ask them what they actually do today and they will give you a laundry list of tedious work:

• Managing calls and last-minute scheduling (e.g., for call-outs).

• Coordinating caregivers with client needs and preferences.

• Processing payroll for dozens or hundreds of caregivers with complex schedules.

• Finding and retaining quality caregivers who can commit long-term.

• Documenting everything consistently and accurately.

• Plus pretty much everything else you could possibly imagine as an operations problem.

There is no default way to handle logistics for these agencies; no existing back office solution that handles the tedious stuff so they can focus on growing and improving their businesses. Existing software, to the extent that it does exist, is legacy and often no better than doing everything manually.

Realizing home care agencies’ burning need for software that could help, Victor Hunt and David Tian started Zingage. The pair, who met at South Park Commons, have slowly built a small, talented team with a single mission: give home care agencies useful new tools to manage logistics so that care can improve for the people who need it most.

Using AI to automate logistics for home care agencies

Zingage has started by solving two of the biggest problems for home care agencies:

• Coordination and ops (like scheduling, intake, billing, and documentation).

• Retention and talent acquisition (finding and keeping great caretakers).

They are doing so with two separate products, Operator and Perform. Operator, their flagship product, is focused on becoming the new back-office for home care agencies. It can do much of the most tedious, but important, work that home care agency owners and their teams are bogged down with: like scheduling, calls, and documentation (among other things). Samantha Tepper, Head of Operations, told us that often the Zingage team will “literally fly onsite to get an agency live with [Operator]. You immediately see the stakes, the workflows being set up in real time, and the impact on their day-to-day operations.”

Victor has described their other product, Perform, as a sort of “Candy Crush for work.” Perform helps caregivers feel appreciated, monitors for signals of burnout, and incentivizes them to do things that are good for both home care agencies and for the people they serve (like accepting longer-term jobs).

Building these products has not been particularly easy; this is not a simple problem to solve.

One example is AI voice agents, which are part of the Operator product. Instead of businesses having to manually make calls to schedule visits, they can use Zingage to make calls automatically. But the traditional batch data approach for these AI phone calls didn’t work—due to the nature of home care, information changed too rapidly. So the team at Zingage built a brand new, more creative data platform for their AI agents that uses a flexible approach to make sure AI agents have fresh and correct data. It works!

(Next, they say they’re focusing on ‘self-healing data’.)

This ingenuity has paid off in a big way: Zingage has reached millions in ARR and already counts more than 400 home care agencies, employing more than 50,000 caregivers, as customers. Their current team, by the way, is just 15 people. And their seed round closed weeks ago. These data points are impressive for a small seed-stage startup. So—how?

The secret to all of this, I think, lies in Zingage’s team and culture.

A high-octane team with a culture to match

Imagine you began a startup job search and were looking for companies with the following criteria:

• Lots of former founders on the team

• A-list talent with experience at great companies

• High-performance, high-accountability culture

• Slow, intentional hiring (the best companies hire slow!)

If there were a piece of software that could show you a list of startups that only match these criteria, Zingage would probably rank pretty high on the list.

For starters, Zingage’s founders have verifiable experience. After graduating from Yale, Victor was the CEO of a real estate software product called Astorian (backed by Bessemer, First Round, and others), which he scaled to more than $50M in project volume. Daniel was the fifth growth engineer at Ramp and one of the first hires for the audience network team at TikTok. Pretty, pretty, pretty good.

One of the founding engineers at Zingage, Greg Meyer, is a former founder who spent years as CTO at Bearworks and as a software engineer at Datadog. Jackie Jiang, who leads design, was previously a founding designer at Tennr. The list, very literally, goes on: there is not a single person on the Zingage team with an ‘unimpressive’ resume. “Everyone feels like a Navy Seal in their niche,” Jonny Gross, Founding Account Executive, told us.

But, of course, resumes only go so far. What matters more is what the team is really able to do. And if there is a through-line in Zingage’s hiring, it is that they have focused on people with proven experience doing real things. People who, in other words, actually do the work. This idea resonated when we chatted with the Zingage team about their culture:

• “Zingage has that magical feeling where you know the team, the founders, the problem and everything is just right. You can really feel the energy, the ownership, and the dedication that goes into solving a real problem that millions of Americans are dealing with every single day. And it takes a very special founding team and great leadership to drive this kind of innovation on an absolutely daily basis.” — Greg Meyer, Founding Engineer

• “This team gets stuff done. I can think of a time where a customer lost their whole administrative team and needed us to step in and onboard them rapidly so we could manage the back office. Ordinarily this is a 4-6 week process. [Instead], we did everything necessary to get this customer up and running within 6 hours, and the results in his business have been amazing. We’re also high agency and egalitarian. You’re expected to come with solutions. And solutions can come from anyone in any department.” — David Knack, Head of Sales

• “Be a pirate with responsibility. Break through walls, then own the outcome fully. Move with velocity, then run cleanup cycles so we can sprint again. Talk to customers. Show your work.” — Samantha Tepper, Head of Operations

• “Company culture is truly a 10. It’s a cliche, but ya, it’s incredible. Leadership has 2050 vision and amazing roadmapping. But interpersonally the culture is amazing. The other week I had a layover in NYC that turned into a canceled flight, posted a message in Slack at 10pm asking if anyone was down to hang, and within an hour I was grabbing drinks with our CEO Victor and talking about product iteration. That’s a dang 12/10” — Jonny Gross, Founding Account Executive

Do a lot of work, do it fast, be high-agency, take accountability. Those are the sorts of things I took away from talking to the Zingage team. And it’s not particularly surprising; you wouldn’t expect a company to do what they have done (millions in ARR, seemingly valuable and complex products, 400+ customers) with what they have (<15 people and very little funding until a few weeks ago) without an intense culture.

Should you consider joining? And if so, how?

When we asked Victor, the CEO, who should join Zingage, his answer was straightforward:

“If you want to be a founder one day then Zingage is the place to learn, if you want to impact millions of sick and senior people then Zingage is the place to build, if you care and want to be proud of your work then Zingage is for you.”

According to Victor, there are “Three Cs for success: CARE, CODE, AND CLOSE DEALS.”

Diet Coke, he says, is also essential.

The rest of the team more or less echoed Victor’s sentiment, noting especially that you need to “high-agency” and that “most days it’s exciting and a ton of fun, [and] sometimes it’s overwhelming.” It is perhaps true to say that Zingage is one of those early-stage startups that is true to how people like Paul Graham write about them: intense, with high expectations and a fast-paced work culture. “You have to know what you’re getting into with an early-stage startup,” David Knack (Head of Sales) told us. “No one is going to [give you] the playbook to be successful in your role.”

You need “grit, velocity, and curiosity” to succeed, Elliott Chang (Product) told us. “In my first interview, they didn’t sell a vision; they shared the next hard problem they were tackling and why it mattered.”

Simonia J., who leads Caregiver Success, says that at Zingage you need to take initiative. “Zingage’s success relies on our shared vision, passion, heart for success. We have limited space for top-down procedure. It really is about what you can actively bring to the table and if you can’t cook, this kitchen is not for you.”

Of course, most startups would tell you that they expect a lot from you, but anyone who has spent any amount of time in tech knows that this is not quite true. There is a sliding scale. Not all companies truly expect the best out of you. (And it may be safe to say that those companies often don’t succeed.)

If you want to be on the more difficult, but likely the more successful, side of that scale, you could consider joining Zingage. Jonny Gross, Zingage’s Founding Account Executive, says you should expect to “work really hard and don’t expect routine.” No matter what, though, Jonny says, “everyone at the company has your back and will support you.

The majority of the Zingage team today is engineers and that does not appear to be changing anytime soon; in fact, the company is urgently hiring 10 engineers. To close, we will end with a few ideas on how you could join (because trying harder is almost always a good thing):

• You could record a Loom video about the most impressive project you’ve ever built, explain why it’s great and why it took so long, and then email it to Daniel. He actually wants you to do this, and says so in his X post. It appears to be the default way to apply :)

• You could actually just build or prototype something you think would be useful to Zingage, noting that you have limited context from the outside, and then email it to them.

• You could write a case study (or tweet, or essay), perhaps like this one, about why you think Zingage is such an impressive company. Then publish it publicly, and email Victor or Daniel a link telling them why you would like to work there and what value you could add.

• You could interview the customers of Zingage’s (few) competitors, collect your findings about their frustrations, and then present them to the team at Zingage.

• You could, as always, write a simple email explaining what value you think you could add to Zingage and why. The more specific the better, and it would especially be good if it is clear you have spent time trying to understand the industry and the problems unique to it.

(Or you could do any of the ideas we write about here.)

TL;DR

If you only have a minute or two, here’s what you should know about Knit.

• Knit is building AI to augment humans, not replace them. Their product makes researchers dramatically more effective. Rather than replacing human intelligence, Knit gives smart researchers more leverage to focus on the most important work. Surveys that used to take months can now get done in a matter of days or weeks, and at a higher quality.

• Solving a $140B+ problem. Running a successful business is, basically, making a series of good decisions. Market research is how many companies arrive at the information they use to make those decisions. Traditionally, it’s been plagued by friction: expensive agencies and months of tedious work. Knit has cracked this code by enabling companies like NASCAR, JBL, Amazon, Overtime, Mars Wrigley, and more to run research faster. Knit may help unlock a bigger market research industry as research becomes faster and more effective.

• Impressive early traction. Knit grew 5X last year, closed a $16.1M Series A, and has quietly assembled a high-performance team—including DoorDash’s tenth engineer, who built the first version of grocery delivery, and Maven’s former SVP of Employer Growth (who helped scale the company from $10M to $100M in annual revenue). They’ve also built a team of researchers so they can better understand their customers and build a stronger product.

• Culture of collaboration and longevity. True to their product, teamwork is a huge part of how work gets done at Knit. People who are a good fit to join are okay stepping out of their defined role boundaries, ready to collaborate, and do not have big egos. This culture seems to have worked; many of Knit’s key employees have been at the company for 4+ years, rare loyalty in a world of early-stage startups where staying for 12 months makes you feel like a veteran.

There seem to be two basic kinds of AI products: tools that aim to replace human output and tools that aim to improve human output.

It’s unclear whether there is something inherently correct or incorrect about either approach. Much of what ChatGPT does, for example, is designed to replace human output—and ChatGPT is quite valuable.

But it is true that humans can be both effective and smart. Our brains, in most ways far beyond anything that has been achieved with LLMs, run on the same amount of wattage it takes to power a dim light bulb. And if today we had to bet between an AI producing an outcome on its own and a smart human using AI to produce an outcome, we’d pick the latter. Humans, as it turns out, are damn good at using tools.

So when we come across a new product that aims to help humans become more efficient—rather than to replace our thinking entirely—we start to pay attention.

Knit is one of those products.

The goal of Knit’s product is simple: to help companies make far better decisions. Knit does this by helping them conduct fast, high-quality market research. Usually that’s a term associated with headaches, piles of money, and three-month slogs to uncover any useful information. Not with Knit.

Using Knit, companies (like NASCAR and Mars Wrigley, some of their customers) can run a research process from start to finish in as little as a week. It’s made possible by AI that gives huge amounts of leverage to talented researchers to do the work they’re best at.

Jeff Herbst, whose firm GFT Ventures co-led Knit’s recent $16.1M Series A, describes Knit’s approach as “augmented intelligence, whereby humans can leverage AI to their maximum advantage… rather than relying on AI to supplant human insights and creativity.” Knit is, as Jeff says, “collaborative [and] researcher-driven.”

There’s a lot to like about this approach; about building products that complement humans rather than compete with them. Mostly there is a lot to like because, in Knit’s case, it is quite effective.

Researchers at JBL, NASCAR, Amazon, and others have already used Knit to do better research, faster. The company grew 5X over the last year and has quietly built an impressive team (their Head of Engineering built the first version of grocery delivery at DoorDash). And in September, they announced their Series A. The goal of the rest of this essay is to pull back the curtain on what Knit is building, why you might care, and what kinds of people may want to consider joining.

Author’s note: At the bottom of this piece, we share some ideas for what to do if you’d like to join Knit.

Market research is valuable, but broken

Imagine an auditorium, a big one, packed out with people who have all bought tickets to a speech you’re giving. The catch? They have no idea what you are going to talk about. There’s an anticipatory buzz in the crowd. You’re about to walk out onto the stage. This is the moment they have been waiting for.

You open your speech with the following words: Market Research.

Odds are, unless the auditorium is packed with people like Aneesh Dhawan, the CEO of Knit, this is not what your audience wanted to hear. And that’s okay! It is okay to admit that market research does not initially get blood pumping and motivation soaring for most people who work in (or out of) tech.

This is odd, though. Shouldn’t market research be one of the most exciting things around?

Running a business is, at some level, just about making a long series of decisions. Whether or not your business succeeds depends on how many of those decisions are actually good. And one of the best ways to make good decisions is to have useful data:

• What would our customers find valuable?

• What would convince non-customers to become customers?

• How would X group of people react if we did Y?

Good market research is how you arrive at answers to questions like these. Which means that in many ways, good market research can be core to making good decisions about the direction of your company.

There are rare opportunities in tech to join a company whose product has a direct impact on some of the most important decisions a company will make. That’s important. Companies think it’s important, too—the size of the market in the United States is currently more than $140B, almost double its size in 2008.

The trouble is that, traditionally, market research has not always been that simple. Instead, ‘market research’ has come with other connotations: tons of money and months of tedious work to uncover information that may or may not be valuable. You often had to hire an expensive market research agency or take up valuable time to research in-house. You had to spend weeks planning and designing surveys. You had to spend weeks on logistics and recruitment. You had to spend weeks conducting the study. You had to spend weeks running analysis, then more weeks coming up with useful insights.

The old way, it might take months and cost ridiculous amounts of money to get results. Which is unfortunate if you need to make decisions quickly (which many businesses do). And which is equally unfortunate if the data you are getting is not accurate and useful (which it sometimes is not).

This is why Aneesh Dhawan founded Knit: to put AI in talented researcher’s hands so they can focus on the highest-leverage work—and so they can go from zero to useful information in as little as a week.

Better information leads to better decisions

Aneesh is a rather interesting guy. In many ways, he checks the classic boxes for what might expect from a great startup founder. At 16, he founded a cause marketing agency called Feed a Friend, whose goal was to reduce poverty and hunger. At 17, he brought solar-powered lights to a school and the residents of a Haitian village. These are, of course, not typical 16- and 17-year old activities.

He describes himself as a “born problem solver” and his team says he “excels at combining humility with conviction.” Angel Mira, an accomplished researcher who now runs customer success at Knit, told me that Aneesh has incredible foresight. When you talk to Aneesh, Angel said, “he’s already considered Door A, B, C, and sometimes D. Impressive.”

Perhaps Angel’s analysis is correct. In January of 2023, Aneesh co-founded Knit with Raahish Kalaria (now Knit’s CTO) to build an AI platform for researchers.

Today, Knit’s product has four main pillars:

• Surveys: Researchers tell Knit the project details and Knit drafts the surveys.

• Samples and Fielding: Researchers can tap into Knit’s network of 65M respondents.

• Analysis: Knit automates the difficult data analysis work.

• Reporting: Knit creates an AI-generated report 24 hours after the data is in.

Finally, if you are a company and would like information before making an important decision, you do not have to wait months. You do not have to sign your researchers (or an external agency) up for hundreds of hours of tedious work. You simply hand Knit to your team and get answers in a week or so.

It’s crucial, Aneesh says, that Knit is built for researchers—and is not a magical black box companies use to get rid of them entirely. “You can run all [your] data through AI, but what’s missing is the context that the researcher brings. We have the researcher at the center of how we run research.”

Momentum is building quickly. In September, Knit announced their $16.1M Series A. They’ve grown 5X over the past year. And, perhaps most importantly, they count more than 60 enterprise companies, like Amazon, NASCAR, and Mars Wrigley, among their customers.

Knit’s enterprise success is particularly interesting. It’s notoriously difficult for startups to sell to enterprises, but top-down sales are important. Failing to sell your product to enterprises can eventually sink entire businesses, but Knit has cracked the code in the early innings. (This is especially important because enterprises spend, by far, the most money on market research. Startups often don’t bother.)

A quietly impressive team and culture

The natural follow-up to all of this is: how? Why is Knit winning so far? The answer, beyond the sheer conviction and effort from their founders, is that they have slowly built a high-caliber team and culture.

Knit’s Head of Engineering, for instance, is Rachel Weems. She was the tenth employee at DoorDash and built the first version of alcohol and grocery delivery, among other things, there. Their COO is Chris Hicks, who was formerly SVP of Employer Growth at Maven (an impressive recent unicorn).

And for their founding customer success role, Aneesh and Raahish made the decision to hire not a career customer success operative but an accomplished researcher in Angel Mira, who has held multiple director-level research roles at his previous companies. This speaks to Knit’s approach more broadly, which has been to “invest heavily in an internal research team” so that they can better understand their customers and build a more effective product.

The resulting culture speaks for itself; or, at least, through the team:

• “From my first conversations with Knit, I was struck by how thoughtful, clear, and fast the communication was. It immediately signaled a high-caliber team that values collaboration and excellence. Now that I’m here, I see that same energy every day. Our ability to move quickly and work seamlessly together is what will make us win.” — Rachel Weems, Head of Engineering

• “I joined Knit for two main reasons. First, the company is at an incredibly exciting moment in its growth trajectory - there is clearly strong product/market fit (as evidenced by the roster of clients that they’ve brought on) and I knew that I could bring my experience scaling startups to have a big impact. Second, I was very impressed by both the co-founders who struck me as competitive and driven, but also humble and curious to seek out diverse perspectives.” — Chris Hicks, COO

• “When I first started we prioritized a hustle culture that was built on being scrappy and being relentless. Today, while we still certainly hustle, we are more thoughtful and discerning with how we spend our time. We make sure that our actions are driving growth for our business in a sustainable way e.g. we are not chasing any and all revenue, but being selective about who we work with.” — Angel Mira, Customer Success Director

• “We prioritize collaboration. We view teamwork as the foundation for success. We play to win. We have a joke saying ‘nobody plays for bronze’ — we are going for gold and always looking for ways to grow and innovate. And we balance high standards with empathy. I love how at Knit we protect and encourage a feeling of belonging while simultaneously pushing the boundaries for exceptional work.” — McKenzie Zajdel, Chief of Staff

Encouragingly, much of the founding team has been working together for longer than Knit has even existed. Logan LeBouef, their Head of Marketing, began working with Aneesh in 2020 on a company that was then called PurPics; McKenzie Zajdel joined back in 2018. (These are people who joined not just before a Series A, but before any kind of major round.) This is a breath of fresh air. At many startups these days, it can feel like staying more than 6 months makes you a ‘veteran’. Not so at Knit.

Creating a more valuable future for market research

If Knit were able to snap their fingers and immediately capture the entire $50B+ that is annually spent on market research in the United States, they would already be one of the most valuable startups in the world. So the market—which is still growing—is already quite large.

Still, market research today is hamstrung by a couple of things. (1) There is a lot of friction associated with doing quality market research, which makes it hard to prioritize and execute at a high quality bar. And (2) there are shrinking market research budgets at many companies; researchers being asked to do more than less. Often, (2) is downstream of (1).

In the best case, market research can be a miracle—it can inform decisions that generate millions or billions of dollars in new revenue. That isn’t the state of most research today, though.

A natural conclusion, then, is that faster and higher quality market research may actually expand the market. More companies would do research, and more often. It would become a bigger part of the business. It would become more central to decision-making.

This is perhaps the future that Knit is betting on. By building the best product for researchers, Knit is aiming to become the default way any company does market research. And, by nature of that product unlocking better insights more quickly, this will lead to a revolution of sort in the industry.

If things work out, Knit will be at the center of how companies discover the data that informs their biggest decisions: about their customers, their products, and more. And, in doing so, Knit will aim to revolutionize the market research industry and unlock many trillions in long-term value via better decision-making.

Should you consider joining? And if so, how?

There are two clear ways to determine whether you may be a fit at Knit: whether you are a fit for the company culture and whether your skills are a fit for their priorities.

We asked people at Knit who should, and shouldn’t, work there. People who are good fits to work at Knit “have NO ego and are team players,” their Head of Customer Success told us, and “are not afraid of building processes and workflows from scratch. McKenzie, Knit’s Chief of Staff, said that good fits are “competitive, curious, open-minded, adaptable, and accountable.”

“Personality-wise,” she said, “we love a good dad joke.”

If you want to operate in a silo, or if you’re happy being complacent (two valid opinions to have!) Knit may not be the place for you. Like many early-stage startups, the culture at Knit is fast-paced and your responsibilities on Day 1 may not look the same as your responsibilities on Day 60. And you will, true to the nature of the product, constantly be collaborating with others on the team. (Again, this is true of many early-stage startups—roles morph and lots of teamwork is expected of you.)

Knit is currently hiring for some key roles that will shape the direction of the company, like Lead Product Designer, Revenue Operations Lead, and a number of engineering roles. This is more or less expected from a startup at their stage building software. What’s unique is that Knit is hiring for a number of research roles; Research Manager, Fieldwork Specialist, Survey Programming, and more.

The opportunity for researchers to work at early-stage startups is somewhat rarer, and so if that is something you have experience in or curiosity about—go check out the open roles.

To close, here are a few ideas from us about how you might join Knit. The goal of the below is not to give you instructions (this is not a guide), but to give you inspiration for creative ways to apply. Because you can often increase your odds of getting hired at a startup if you put in more effort than others.

• If you were applying to the Research Manager role, you could speak to 5 UX researchers and gather insights on what’s important to them, and then make a Loom or presentation showcasing what you learned and how you’d use it to increase retention at Knit. (You could caveat that you don’t have too much context).

• If you were applying for the Founding Full-Stack engineering role, you could create a custom interactive website—made just for the Knit team—showing off the most impressive projects you’ve built from 0 -> 1 and explaining why you think they make you a good fit to work at Knit. You could then email this along with your application.

• If you wanted the Product Designer role, you could design an interactive prototype for a new feature you think would be useful to Knit. You could even include custom branding as if it were a demo designed for a specific one of their customers (like NASCAR).

• If you wanted the Revenue Operations Lead role, you could do your best guess at estimating their current sales process and suggest a few ways to reduce sales cycle time for enterprise deals.

• For any role, you could spend an hour researching Knit, then write a memo explaining the specific ways in which you think you could be valuable. The more specific your ideas, the better.

• For any role, you could pay for a survey to be run (or run it yourself) on potential Knit customers asking some question that would be useful to Knit, then email the findings to Aneesh along with an introduction of who you are and what you do.

• You could also, though this is lower-effort, just email someone at Knit in addition to sending in a more formal application. Here is some advice for emailing.

(More creative ideas for getting hired at startups here.)