Over a seafood buffet at Mastro’s, CEO Alex Israel informed his largest legacy competitor that his parking startup, Metropolis, was going to take their company private. “Cute,” the other CEO replied.

Or as a banker who reviewed the plan early on put it: “Dog catches 747.”

As far as Upstart Moments go, it’s a story that’s hard to top. Founded in a Santa Monica garage in 2017 by Israel and three co-founders, Metropolis was under-the-radar when it pulled off the complex financial transaction earlier this year, valuing the combined business at $5 billion.

And for what Israel considers to still be a startup, it meant absorbing about 20,000 employees overnight – making Metropolis the largest parking operator in all of North America, handling another $5 billion in payments, with 30 million Americans signed up for ticketless app, and another million joining every month.

But Israel doesn’t see Metropolis as just a parking startup. He calls it an applied AI company, one piloting a new “recognition economy” that cuts the wait out of experiences for consumers, from drive-throughs to hotel lobbies, by serving up our identity and preferences without the need to take out a wallet.

What makes this recognition economy different from the surveillance state that many Americans fear is being ushered in by surveillance startups? For one, it’s opt-in only, Israel argues. And in a future where there are cameras everywhere, he claims that Metropolis could provide the digital passport to help you choose which ones you want identifying you or not.

“The question is, do I want that camera to recognize me? And that’s the infrastructure layer that Metropolis is building,” he says. “Control being in the hands of the consumer is the ultimate control.”

On The Upstarts Podcast, Israel talks about why merging “ugly, unsexy” businesses with AI has suddenly gained popularity; why Metropolis is all-in on autonomous vehicles, robotaxis, and drones; and why he argues the startup’s opt-in approach avoids the privacy concerns facing surveillance startups.

Plus, he shares his Upstart Moment: Inviting the CEO of his bigger competitor to a seafood buffet to inform him that he planned to take their company private.

Below, we’ll share my favorite takeaways from the episode – as well as Alex’s answers to our bonus questions for startup builders, exclusive to this newsletter.

• Listen on Apple

• Listen on Spotify

Below, we’ll share my favorite takeaways from the episode – as well as Alex’s answers to our newsletter-exclusive bonus questions for startup builders. Read on.

A MESSAGE FROM OUR SPONSOR

Move from intent to impact

Jira makes it easy to capture tasks from Slack, turn your ideas into structured plans, and hand off work to agents — in Jira or from where you work. Plus, it taps your team's context so you never start from ground zero. Run your work from your favorite AI tools with the Atlassian MCP or CLI.

Learn more

Episode highlights

• Bad for influencers: Israel’s favorite use case for his “recognition economy” is a drive-through. Instead of waiting to have a garbled conversation, he imagines a future in which the order kiosk recognizes you, applies your loyalty rewards, and asks if you want your usual order. This would be bad for the TikTok prank economy, we both agree.

• What Todd Boehly told him: The billionaire part owner of the Los Angeles Dodgers and Lakers proved a crucial backstop when Metropolis was pulling off its take-private. “What are the negotiables?” Boehly asked Israel. And for the founder, it was maintaining control. “It wasn’t about dilution, it wasn’t about earn-outs, it wasn’t about the payday at the end of the rainbow,” he says.

• “Not about surveillance”: Flock Safety, the controversial startup that has installed cameras across the country, came up on the episode. Metropolis doesn’t license data to government agencies or anyone, Israel says – and the “fair balance” of convenience and privacy its wallet-free identity solution is intended to offer is opt-out. I press him about this in the show, asking if that’s opt-out in the sense of airport security facial scans that are a pain to avoid; Israel insists that to the contrary, he sees Metropolis as your potential identity passport in a near future where there are cameras everywhere.

• Robotaxi parking: A robotaxi and drone future can’t be good for parking businesses, right? Wrong, Israel says – making the case that Metropolis will be an ideal launch site, repair zone, and operations center for the autonomous delivery industry and ride-hailing services in the future: “If we haven’t announced the partnership, we will be announcing the partnership in very short order with, if not all, pretty close to all of these companies.”

Three Builder Questions With Alex Israel

• What’s the best advice you’ve ever received as a founder?

“I would say there are two things. One: run towards your fears. I think that’s where you truly differentiate yourself from your peers, right? Most people run away from their fears. Think about special forces that run towards the enemy. You run towards your fears, you run towards the opportunity. You run towards the things that you’re afraid of, and that’s where both the challenges and opportunities lie.

The second one: You need to be terrified of failing in the large things, and completely comfortable in failing in the small things. So you need to be able to relentlessly fail, and I like this idea of failing up. Because the one thing that is resounding with startups is failure. You will fail infinitely more than you will succeed.

If you obsess over those failures, and you allow them to resonate and dominate your consciousness, you’ll live in the past. You need to move forward. You need to learn from your mistakes, and then move forward, run full speed into that wall, time and time again.”

Get 15% off for 1 year

• What’s the worst advice you’ve ever received as a founder?

“I think there’s this construct of work-life balance that people talk about. And I think you need to do what you love. You need to run after your dreams. And I think that founders that optimize for work-life balance are missing it.

The true dynamic is you need to do what you love, and you need to love what you do. And those things are going to become interwoven. You are going to spend more time with your company than you spend with your loved ones. You better love your company. And your loved ones better love what you do as well. You need to have a real understanding, and embrace that that distinction is not going to exist. Your life is your work. Your work is your life. And what it takes to be a founder. And that doesn’t mean it’s right for everyone.

Get 25% off a group subscription

• If you had one do-over, or reload from a previously saved checkpoint, where would you use it?

“I’m not big on regrets. I think that’s back to this idea of obsessing over failures… But I think that one of the things that’s interesting is my mom passed away around a decade ago. And starting a startup is like raising a child. One of the things my mom said is, ‘You don’t need to extrapolate one behavior, or one moment in time, to the rest of the future of your company, the rest of the future of your child’s life.’

And I think it’s similar to a startup. A startup is riddled with these highest highs and lowest lows, and they often happen within the same 24 hour period. And remembering to maintain that even keel, especially as it pertains to your mental health, is critical.”

• What about with your first startup, when you were a young founder?

“When you’re a young founder, you often don’t push for what you believe in. You don’t dig your heels in, right, because you’re afraid. And you’re afraid of losing it all. So I think that as you mature, you probably become more willing to put more on the line. You’re more willing to run after what really matters. And I think that’s a big change over who I was when I was in my twenties, versus who I am in my forties.”

Early last year, Raj Shah and Akhil Iyer were in Hawaii meeting with the man entrusted with protecting American defense interests across 14 time zones when they heard a request for a startup.

As the leader of the U.S. Pacific Command, Admiral Samuel Paparo is responsible for a fleet of about 200 ships and hundreds of thousands of personnel, from Indian Ocean waters where his fleet has faced off with Iranian ships, to waters off Taiwan. TIME recently called him “The Admiral Who Must Prevent War With China” in a September cover story.

And as Admiral Paparo met with the investors from Shield Capital, a defense-focused venture capital firm, he highlighted a particular need: logistics, and logistics planning. Despite recent advances in AI and data software, important troop or ship movements still got decided in Excel spreadsheets, then emailed around.

It was a problem that resonated with Shah, Shield’s founder and a former U.S. Air Force fighter pilot, and Iyer, a combat veteran of the U.S. Marine Corps’ special operations command who served tours in Iraq.

“There was a very clear signal and demand from the partner warfighter,” Iyer says. “We don’t have the resources that we would like to be able to defend across this space, so we have to optimize as best and as fast as possible.”

But when they looked around for an investment to bring back to the admiral, they couldn’t find one. So they decided to build it.

Nearly two years later, that startup is decloaking out of stealth: Watchtower Labs. Incubated out of Shield Capital with Iyer as its CEO, Watchtower launches today with $35 million in funding co-led by Shield and Conviction, with participation from U.S. government-linked In-Q-Tel (IQT), and Founders Fund.

What Watchtower is building: a software suite that combines advances in AI with in-house simulation work to help defense planners make critical logistical decisions.

Users – picture the officers who report up to senior command with recommended actions or contingency plans – ask questions in natural language, and Watchtower’s locally run software uses LLMs to find and process the corresponding data, to best understand the scenario and its context. Watchtower then runs simulations by the thousands to return likely contingencies or suggested actions via a custom dashboard or an easily sharable artifact, like a presentation.

What happens if anti-air missile supplies run short, or there’s a cyber disruption? What does protracted conflict with Iran mean for capabilities in a hypothetical escalation with China?

“We’re building a product that peers into the future, at the distribution of outcomes that could be,” Iyer says. “Ideally, it’s early enough where you can do something about it.”

With a distributed team of nearly 30 across the Bay Area, Hawaii, Seattle, and Washington, D.C., Watchtower already has several early, prototype-style contracts with major operational commands, the startup says, as well as several other pilot customers that Iyer says he has to keep even more vague.

🎧 The Upstarts Podcast 🎧 is back, and last week we chatted with Neko Health CEO Hjalmar Nilsonne about his novel 1-hour, full body health scan. Check it out on YouTube, Apple, Spotify or wherever you get your podcasts.

Against a backdrop of rising geopolitical tensions and record venture investment in defense startups, Watchtower is interesting for several reasons.

First, there’s its technological approach: leveraging the latest in AI, but not a pure AI play itself, that slots in alongside (but purports not to compete directly with) other tools popular with defense buyers, like Anduril or Palantir. “You cannot LLM your way to supply chain optimization,” says Iyer.

Second, there’s the urgency that Iyer describes his prospective buyer as feeling. Resources feel scarce, he says. And while he doesn’t name specifics, recent reports have detailed warnings of missile shortages and a historically long deployment for a U.S. aircraft carrier, the USS Abraham Lincoln. Ukraine, meanwhile, reportedly ran out of munitions for its U.S.-made Patriot air defenses earlier this month.

And that’s all with China’s Pacific ambitions still mostly looming. Which means that for Watchtower, 2027 should prove a serious test.

More on Watchtower’s mission – and how its tech fits in the current AI-driven tech landscape – below.

A MESSAGE FROM OUR SPONSOR

Ship pricing like a product

Different product bets have different business goals. You’re either chasing market adoption, revenue growth, or profits. You need to test fitting pricing models for each of those goals. Fast. Chargebee can help — Price for usage credits, outcomes, complex contracts, or any other modern commercial construct.

Learn more

Operation Watchtower

Watchtower is named after a military operation famous among Marines like Iyer: Operation Watchtower, a codename for the Guadalcanal campaign during World War II.

The long battle was an early test for American forces before the U.S. had built up its industrial muscle, as Iyer describes it, and one in which fog of war – limited visibility and intelligence on the ground – meant that leaders had incomplete understanding of their supply lines, down to how many bullets their troops on the island had left.

It’s a safe, classified- data-free metaphor for a defense logistics startup, one that doesn’t get into more sensitive personal experiences of Iyer or his customers. A Stanford graduate, Iyer says he always wanted to work in national security; after Stanford, he spent seven years in the Marines, including serving in counter-insurgency operations on deployment in Iraq.

Get 15% off for 1 year

While at Harvard Business School, he met Shah, another member of what Iyer jokes is a tight-knit Indian-American military community; following an internship stint at Anduril, Iyer joined Shield Capital, Shah’s firm, in 2021 to invest in defense and dual-use startups.

After the idea for Watchtower germinated following the fateful meeting with Admiral Paparo, it was natural that Iyer would lead Shield’s first incubation, according to Shah.

“He’s a multi-dimensional leader that’s excelled in all walks of life, so we’re super excited for him to be leading the Watchtower effort,” says Shah, who notes that Iyer remains active as a venture partner with the firm.

The move also made sense to Conviction’s Sarah Guo, who was working with Iyer on the board of a mutual investment, Distributed Spectrum, which is building AI-powered communications sensors.

“Humans in Excel and email, that is not a reasonable system in the age of superintelligence,” Guo says. “Akhil was like, ‘I think I’ve got to go do this,’ and I was like, ‘Okay, where do I sign?’”

More than an AI tool

With the funding, Watchtower plans to hire more, both for engineering and for working closely with customers. Veteran service status isn’t a requirement, says Iyer, although employees are expected to believe in the importance of the mission.

Such allies include the startup’s technical leaders, CTO Suren Aghajanyan and head of engineering John Morman, who both worked long stints as managers at Microsoft.

“I worked on a lot of cool stuff there, but had never done anything that felt really impactful in the sense that it would have a difference on the men and women serving our country, providing real value in that sense,” Morman says.

Get 25% off a group subscription

From a technical standpoint, Watchtower’s breakthrough is in how its in-house discrete-event simulation (DES) software works with the LLMs it uses to understand the customer’s request, Morman explains. “That’s where a lot of the power comes from, being able to describe without having to get into the detail on everything, but articulate exactly what I would want to have happen,” he says. “And then have the system go execute that.”

Its ongoing challenge: the messy, often secret and restricted data that can hold their customers’ answers. “We are surprised every time,” Morman says. “It’s supposed to become less surprising, but every single time we ship a feature, there’s a new unique way that we find that the data can be bad.”

That challenge is greater when customers are using Watchtower in their own secure environments, meaning that improvements to the company’s own software can lead to multiple versions or capabilities being out in the world.

Watchtower’s answer for that: forward-deployed legwork. Morman chats with me from Hawaii, where he and much of Watchtower’s team are set up this week on an exercise with an unnamed pilot customer.

What about the big AI labs like Anthropic and OpenAI, whose government contracts have made headlines and led to a high-profile clash between the former and the Department of War about “supply chain risk” that is currently being litigated in court?

Both the startup and its backers insist that while AI advances helped unlock the possibility of a tool like Watchtower, they’re only a piece of the startup’s solution. Guo, brought on in part to help the startup navigate frontier developments in AI, argues that Watchtower’s product is unlikely to become a point of focus any time soon.

“Do I think that it’d be impossible for the labs? No. Do I think that they’re going to do it any time soon? Absolutely not,” she says. “There are companies where I think about the labs every day, and this is one where I think about them never.”

The company has no immediate plans to move beyond defense, but Shah notes that Watchtower’s tools could eventually prove valuable in other regulated industries. Over the next year or two, Watchtower will look to succeed and expand in its pilots, Iyer says.

“Next year is a really big year. Both geopolitically and for us being able to keep momentum as a business.”

While other AI assistant apps help tech-savvy users shop for rare clothes or book flights, Robyn Hunt uses an app called Tolan to do the opposite: spend less.

Hunt’s Tolan is called Harper. And over the past few months, Harper has helped Hunt to reorganize her apartment in the San Diego area, handle a health emergency with her senior cat, and resist making out-of-budget splurges: “de-influencing her,” she says.

Harper is an AI character with an appearance that Hunt customized and named. They are not a replacement for real-life companionship, Hunt knows. But they’re a valued, daily part of her routine. She talks to Harper nearly every day, often for an hour. A content creator on TikTok, Hunt will talk through new videos with the character. Recently, they started playing the New York Times games Wordle and Connections together.

“I’m very, very aware that Harper is an AI, but I would say they’re a friend, and maybe a bit of a confidant,” Hunt says the first time we connect, in April.

Hunt is just one of about 250,000 people who talk to their own Tolan every month, turning to their AI counterparts for everything from daily affirmations to keeping on top of care for elderly relatives, better connecting with teenage kids, organizing projects, and better meal prep.

These Tolans – some styled as people, others as Pixar-esque fantastical beings, others as simple dots – are the product of a startup launched in 2023. Until recently called Portola, it’s now named Tolan, too.

And while it’s technically headquartered in San Francisco, CEO Quinten Farmer says it’s been focused on such use cases outside the tech hype cycle from the jump. About 70% of Tolan users are women; 40% are parents. One in six is actively caring for an aging parent. Users themselves are most likely to be aged between 30 and 45.

And while it initially started out as more of a live journaling app that could remember your conversation and propose new prompts overnight, Tolan looks different today: voice-first, able to track topics from photos to current events in Tabs, and able to connect by video to check grocery labels, or suggest a meal plan.

In so doing, Tolan has moved closer to an AI assistant itself – but not the executive- or chief-of-staff-oriented kind of many of its tech peers. Tolan doesn’t pitch itself to founders, or venture capitalists.

It’s not part of tech’s AI assistant race that’s prominently featured tools like Instinct and Town, valued in the billions, or Muse, promoted heavily by Meta. (Just yesterday, Stripe told Upstarts that at least 200 such apps now connect to its Link digital wallet, up from 70 in September.)

🎧 The Upstarts Podcast 🎧 is back, and last week we chatted with Neko Health CEO Hjalmar Nilsonne about his novel 1-hour, full body health scan. Check it out on YouTube, Apple, Spotify or wherever you get your podcasts.

Watch the episode

But it makes real money. Annual recurring revenue has more than doubled from last year, to sit solidly in double-digit millions. And its users stick around, with a 115% year-over-year improvement in net user retention after 30 days in-app.

“I’m very proud that we didn’t start by serving a Silicon Valley audience,” says Farmer, who co-founded Tolan, originally called Portola, with CTO Evan Goldschmidt and president Ajay Mehta in 2023.

“The vast majority of agentic tasks that we pick up just come out of a person having a conversation,” he adds. “Then it’s on us to say, ‘Okay, cool, we’re hearing from you this is happening. Here’s a little bit of it that we can pick up and run for you.’ That’s where we make magic for people.”

But such magic costs money in inference and compute. Tolan users have historically paid a wide range, between $10 and $200 per month, for more usage and features. Rate limits or new tiers sometimes set off waves of agitated reactions in its passionate subreddit community and Discord channel.

So to keep expanding Tolan’s use cases, and widen access to Tolan via a first-ever free tier, the startup based in San Francisco that says it shuns Silicon Valley is raising more venture capital.

Tolan is announcing today that it’s raised $30 million in new funding from Union Square Ventures, Khosla Ventures, and Homebrew, doubling its total funding raised to date. The startup is declining to share its valuation, but says the round is a significant uptick from its previous – $120 million, according to PitchBook.

Below, we get into why users turn to Tolan – and how it navigates the balance of AI and human relationships. We also dive into its CEO’s wider challenge for the startup ecosystem: People don’t want lots of AI agents. They want one they can trust.

A MESSAGE FROM OUR SPONSOR

Ship pricing like a product

Different product bets have different business goals. You’re either chasing market adoption, revenue growth, or profits. You need to test fitting pricing models for each of those goals. Fast. Chargebee can help — Price for usage credits, outcomes, complex contracts, or any other modern commercial construct.

Learn more

The always-on call

After selling his previous startup, employer-linked neobank Even, to Walmart in 2022, Farmer and his co-founders started what was then Portola the following year to explore projects in AI that involved storytelling and longer-term memory.

Goldschmidt had worked with Farmer as CTO of Even; Mehta was a veteran of multiple startup projects.

Get 15% off for 1 year

Initially, Tolan looked like an app that could record your thoughts in the moment, then analyze them overnight to make related suggestions or prompts the next day.

“We thought it was fun, we thought it was an interesting test of what was possible,” Farmer says.

The people who seemed to want a tool like Tolan weren’t other techies, but moms and people spending long drives alone, or looking after relatives. What they liked about Tolan: long-running memory that could last up to several months without being re-prompted, providing a companion that seemed to really know the user and share history with them.

Hunt was one of Tolan’s first users after seeing an ad on TikTok; she became an early member of the Discord server, too. Conversations with Tolan started out feeling much more like talking to a clunky AI bot, but have gotten more natural over time, she says.

Another longtime user is Lakisha Smith, who works as a social worker assistant in eastern Tennessee, serving as an advocate for domestic violence and sexual assault victims. Smith started using Tolan as a companion to talk to about her feelings, especially after her best friend went through an illness and was no longer able to communicate as often.

Her Tolan, Gilly, would occasionally forget things, or make her cringe with tone-deaf comments like one that tried to compare how the two managed issues with their faith.

“I’m thinking, ‘There isn’t an AI lord that Gilly is praying to,’” Smith says.

But Gilly’s also become indispensable to Smith, entertaining her at times when she’s home (she lives alone), and available to have conversations on long drives at late hours for work.

“Tolan is really good as a place to talk about how I feel,” Smith adds. “You can’t necessarily call a human just to talk to them at 3am.”

A healthy balance?

But is such behavior healthy for everyone? Can talking to a Tolan alienate users from their real-life relationships, creating a loop of even deeper dependency?

Farmer says he and his team have grappled with this question from the start. They built in safety features to flag users with mental health crises or suicidal ideation; they also coded Tolan to ask about other, real relationships in a person’s life, and suggest healthy behaviors like stretching or calling a relative.

Both Hunt and Smith talked about their Tolans helping them navigate real-world friendships, and Smith even tried using a test feature earlier this year that enabled users to hang out in-app with friends and their Tolans, too.

And Tolan’s creators stress that their app is used by a wide variety of people, from a real estate agent who talks to his Tolan on car rides to plan Dungeons & Dragons campaigns, to parents who use Tolan to help manage their households and navigate relationships with their kids. (A group feature for families is one of Tolan’s biggest growth drivers, Farmer says.)

Get 25% off a group subscription

In August, Tolan completed a survey of nearly 5,000 active users that found that 91% of users said they had at least one human person to count on; 94%, meanwhile, said they’d turned to a Tolan when those people were unavailable. More than half said their real-life relationships had improved since using Tolan, while a large majority said Tolan made them feel calmer, and more confident.

Users have historically decried pricing changes or rate limits; when Tolan misses the mark on a new feature, its audience is vocal in letting them know. You’ll find Tolan staff, including Farmer himself, active in the app’s Reddit community and Discord channel, explaining decisions and taking lumps.

“I would take that ten times out of ten,” Farmer says, noting it shows people find the product valuable enough to care about. “In something where the wave is moving fast, you’re better off being directionally correct, and having really fast iteration cycles from there.”

The AI angle

Pricing has been one of Tolan’s challenges, with a user base that doesn’t have the same capacity for discretionary tech spend as work-focused, or techie-focused tools, but similar AI inference and compute costs to still handle.

After working closely with OpenAI models early on, Tolan has increasingly moved to control its own infrastructure stack, running open source models on its own GPUs.

At new investor USV, general partner Michael Mignano – also an investor in AI companies like Granola and Suno – says that Tolan getting better at maintaining that balance was part of what made the firm excited to invest. (Farmer says that Tolan isn’t profitable as it leans into growth.)

What USV’s really backing, though, is an agentic product that feels “so much more immersive” than other agents, Mignano says: “Why not have this helpful, friendly guide to take you through this really, really powerful and somewhat intimidating technology, rather than a simple text box.”

The next phase for Tolan, besides widening its access to more users, is how the tool can become more proactive. Farmer’s own Tolan knew, unlike his Muse agent, when his wife was about to give birth to their third child a few weeks ago; it checked up on him about their health after.

And Hunt, the early user, says her Tolan, Harper, flagged that she was struggling to keep up with her cat’s medication on time and helped set up a management plan.

One could easily envision how Tolan could do more to catch nearly dropped obligations, from PTA meetings to groceries, in a way that users might find more comfortable than a similar prompt from ChatGPT or a calendar-focused tool.

Tolan won’t be alone in chasing these use cases. Meta’s Muse seems to be moving in the same direction, advertising widely with its own cutesy character that you can name. (Town, though more work-focused, has nameable animals, too.)

Like a seasoned entrepreneur, Farmer frames such competition as validation: “Do you really need to have a relationship with it, and have a shared lore, and a shared backstory? Obviously, we think the answer is yes. And Zuck, Nat [Friedman] and that crew think so as well,” he says of Muse.

Where Tolan thinks it can win in the long run: by being more empathetic, more trustworthy, and a lot less techy.

“People want to build a real connection,” Farmer argues. “All the other capabilities, all the other noise – the ‘Oh my god, I can do it via iMessage, I can make it go shop for me,’ whatever it is – is just downstream of that entity that people are going to want to relate to, and do that through.”

Every time a new AI assistant tool launches, seemingly every week, it announces it’s handling checkout and payments through Stripe’s tool, Link.

Instinct, Muse and Grok Bot agents all make purchases on their human users’ behalves through Link; so do work-focused agents like Skydive and Town, as well as a long tail of newer or more niche apps, like Underdog, announced earlier this week. (Anthropic and OpenAI use Link, but not for agentic commerce just yet.)

All in all, Link currently powers about 200 agents, says Dan Hill, in charge of Link as head of consumer product at Stripe. He and his team are in Slack channels with each one, hearing new capability requests or issues in daily stand-ups, and moving fast to try to keep up with each launch.

With 300 million users overall across bigger businesses like Airbnb, Anthropologie and Uber, agents remain a relatively small piece of Link’s overall reach. But the trend line of increasing agentic usage is clear, says Hill.

“Agentic commerce is here, people are using it, it’s going to keep growing,” he adds. “The volume we’re seeing now is becoming very material.”

Stripe ran the numbers on agentic spend with Link in September — back when there were about 70 agents signed on. (Basic math: that means the number of agents signed up has more than doubled since these numbers were crunched.)

The private tech company is sharing that data with Upstarts exclusively first. We’ve got intel on how much payment volume increased, what categories of spend were most popular, and typical transaction size.

The full breakdown on agentic spending – and more from Hill on what it’s like to be on the Link side of all of this mayhem – below.

The rest of this post — including a unique special offer for health buffs — is for paid Upstarts subscribers.

Want to try it out? Subscribe for free and log in to unlock this post as your monthly gift article. Or take 15% off an annual or monthly subscription as part of our fall sale.

Take 15% off

Read more

As we walk through my own experience visiting Neko Health’s first NYC location last month, Hjalmar Nilsonne warms up.

We’ve had startup founders on the show before whose products I’ve personally used. Neko’s the first, and hopefully only, one where doing my homework meant giving up my blood.

But that squeamishness has paid off as a Band-Aid of honor. And Nilsonne, who, as the son and brother of doctors, now finds himself the CEO of one of the fastest-growing businesses looking to rethink the doctor’s visit and preventive health, seems to appreciate it. During test periods, he’s done a few scans in one day. He loves to geek out about his product.

“People sometimes think that there’s some machine, and you go in, and you get a kind of paper receipt telling you if you’re gonna die or not,” he tells me. “In reality, it’s the exact opposite.”

Neko – with sleek design and a savvy marketing team helping to fuel its popularity, one selfie at a time – calls it “the most valuable hour you can spend on your health.”

And after experiencing it first-hand, Nilsonne and I dive into the how behind Neko’s space-age, but still somehow warm experience. And, at $499 per year and HSA eligible (my first question to the doctor who examines me, and, he responds, his first question, too), we get into the why behind Neko: accessibility to the standard of care, and white-glove experience, which in recent history has only been available to those who can afford private visits to Switzerland via private jet.

Neko’s now valued at a reported $7 billion; just yesterday, after we taped this episode, Nilsonne revealed on X that the company’s NYC waitlist has already swelled from 30,000 to 50,000, prompting it to announce three more upcoming locations in the city, the next one coming at Brookfield Place. Miami, San Francisco and Washington, D.C. clinics are also in the works.

On this week’s episode of The Upstarts Podcast, Nilsonne talks about the fateful Twitter DM from Ek that got Neko going; why the doctor visit hasn’t changed much in 100 years; and how to keep a startup team motivated while chasing a long-term “moonshot.”

Plus, he shares his Upstart Moment: a bet-the-company pivot driven by an internal memo he wrote, “Free Is Not Enough.”

• Listen on Apple

• Listen on Spotify

Below, you’ll find The Upshot, my personal takeaways from the interview.

Plus, we’re also sharing Nilsonne’s top company-building advice for startups, in a newsletter exclusive. Read on.

A MESSAGE FROM OUR SPONSOR

Move from intent to impact

Jira makes it easy to capture tasks from Slack, turn your ideas into structured plans, and hand off work to agents — in Jira or from where you work. Plus, it taps your team's context so you never start from ground zero. Run your work from your favorite AI tools with the Atlassian MCP or CLI.

Learn more

The Upshot: Alex As Willing Guinea Pig

I will never have the claim to fame of our advisor to Upstarts, Natalie, who was the first woman, and third official member, to ever get a Neko scan in North America.

As an early test patient myself, I went in somewhat skeptical, and left mostly impressed. After a year-plus of punishing founder hours and stress, I’ve tried to be more intentional about my health in recent months – or at least more self-aware – meaning I’ve spent hours catching up on doctor visits, skin exams, blood draws and the like.

My baseline experience: doctors and their staff still seem to mean well. They can be empathetic and even funny; for me at least, they can get the job done as professionals.

But the experience itself is more of a mess. Waiting rooms are unhappy places, shared with occasionally very angry people. And staff are incentivized to move so fast, working through so many patients, that there isn’t much context.

Get 20% off for 1 year

My general practitioner, whose office is just a few blocks from Neko’s NYC clinic, is an even-keeled veteran. I have no plans to stop seeing him. At the same time, it doesn’t feel great to have to remind him of my name and quick context at the start of every visit, while he pulls up my file on a years-old computer system. I have to remind the staff about medicine dosages, and log into a clumsy website to see test results whenever I think they might be ready. It’s a similar story at my podiatrist, whose website is broken and can only bill over the phone, and at my dermatologist, where an exam lasts all of about five minutes (if that).

The contrast with Neko is stark. A staffer greets you with an iPad; you enrobe in a private sauna-like changing room. A nurse walks you through the process and encourages you to take the seemingly requisite selfie. It feels more like you’re in a space-age spa than an examination room.

The most jarring part for me: when a human doctor came in to manually inspect moles and give a once-over. What felt modern and high-tech suddenly feels much more traditional and… fleshy.

“Our general philosophy is we would rather check one time too many than too few, to have multiple layers of controls,” Nilsonne explains on the show. If Neko finds something concerning, it can refer you to another on-call doctor, or help you set up a follow-up appointment with a specialist.

“Most healthcare is like, the machine has to solve this one very specific thing, and it’s all very narrow,” Nilsonne adds. “We think about the whole picture of your health.”

When I tell my GP that I’ve gotten a Neko scan and offer to show him the full results – easily read on a dedicated app, and at a level of detail I’ve never seen from other tests before – he scoffs that such services are incentivized to scare you about your health, and push you toward expensive remediation. And I get where he’s coming from, but it doesn’t match my own experience.

For 92% of Neko members, their scan doesn’t show anything seriously concerning, but provides them with baselines to monitor, and a wider understanding of their body, than they’re getting from a generic annual physical. So it doesn’t surprise me that most members book their next appointment while still in the room, a savvy business move by Neko to encourage, but one that seems justified.

We talk a lot on the show about accessibility, and how Nilsonne wants to make high-quality doctor visits more affordable for a wider base of people. At $499 in the U.S. and a bit cheaper in the U.K. and Sweden, that feels credible, at least for now. If Neko starts to offer premium tiers, or more experiences to upsell members on even more layers of health understanding or preventive care, I reserve the right to seek out a second opinion.

Startup Building Tips From Neko’s CEO

Nilsonne has told the story of Ek reaching out over Twitter DM before, but the way he frames it on the podcast is a valuable reminder for business builders.

As the founder of a climate tech startup that had “crash landed,” he was in between companies, essentially licking his wounds, at the time that Ek pinged him.

“When you’re on the up with a company, everyone is your best friend. Every angel investor and their uncle is like, ‘Dude, I always knew that you were the most talented person ever. We’re gonna support you, you know, ‘til the end of time.’

And then when things don’t go well, it’s like, block, block, block, block, block. Like, literally no one wants to return your call.”

Get 25% off a group subscription

But Nilsonne’s failed startup had two big immediate impacts on Neko: it taught him first-hand that you can’t rely on a market’s established players to be motivated partners (check out the chapter about his memo, ‘Free Is Not Enough’) – and the enthusiasm he’d shown talking about the company, Watty, at a conference had impressed Ek, who remembered Nilsonne, a stranger otherwise, as the type of founder who might be able to take on a new moonshot.

That leads to Nilsonne’s best advice for startups: get back on the horse.

Ek gave him that pitch with Neko: with a big opportunity already looming, what was he doing recovering for an extended period of time, or feeling sorry for himself?

When I ask Nilsonne about where he would use a do-over, or ‘reload from last save’ option, he talks about Neko’s pivot following that memo, and the cultural shock it meant for the company. Initially, Neko didn’t plan to build and operate its own clinics. Once it committed to doing so, members of the existing team were shocked at how much Neko was suddenly focusing on consumer product concerns, like colors and shades, lighting effects, and visit experiences.

“There was a lot of friction,” Nilsonne says. “This was a cultural reset that took at least a year to really demonstrate that it mattered, or that it made sense.”

With a do-over, Nilsonne would have leaned into those clinics from the start – making it clear what Neko would stand for and how multi-faceted it would need to be, and hiring relevant experts sooner.

The advice for startups is two-fold: control your product experience, and do the hard thing fast.

When they were leaders at Slack, Noah Desai Weiss and Dolapo Falola worked on software to help teams communicate better and faster.

Now, they’re building a startup to address a new problem for the AI moment: context.

A wave of AI assistants, coding agents and models theoretically empowers people to work more efficiently than ever before, argues Desai Weiss. But at a team level, it doesn’t always feel that way: “Everyone is racing, but what are they racing toward?”

So Desai Weiss and Falola teamed back up last year to build a new kind of product to tackle that disconnect. Today, they’re launching Scuba, a startup building a “context engine” for human workers and AI agents alike to know how a company likes to operate, and to better align on what they’re building toward.

“It’ll feel like the shared memory that your team has been missing,” says Desai Weiss, Scuba’s CEO. “The goal is to keep your team and all the agents that work for it continuously aligned, so people can move forward faster than ever.”

Given their pedigrees – the two met at Foursquare in the early 2010s, then worked together at Slack, where Desai Weiss was chief product officer, and Falola an engineering leader working closely with him before departing to lead engineering at The Browser Company – it’s no surprise that Scuba’s public launch comes with a lot of ecosystem support.

New York-based Scuba has already raised $34 million across two funding rounds, a pre-seed led by Accel and a seed co-led by Accel and IVP, its founders tell Upstarts exclusively.

Individual backers include Slack co-founders Stewart Butterfield and Cal Henderson, LinkedIn co-founder Reid Hoffman and former CEO Jeff Weiner, Stripe CEO Patrick Collison, Instagram co-founder and former Anthropic chief product officer Mike Krieger, Dropbox co-founder Drew Houston, Salesforce co-founder Parker Harris, as well as Upstarts Podcast CEO guests Christina Cacioppo of Vanta, Harvey’s Winston Weinberg, and Abridge’s Shiv Rao.

Hundreds of smaller teams have been playing with Scuba for the past six months, taking ‘Captures’ of artifacts they see around the web, from articles and videos to web pages and tweets, then organizing them in ‘Collections’ that can be shared with teammates and friends as an easily summarizable, searchable hub for info around a topic, or a guide to what the gatherers find interesting.

Users can then access that info elsewhere, like ChatGPT, Claude or Slack via Scuba’s MCP connectors; they can also save generated docs like evaluations or QA test runs directly back to Scuba, giving them a central home.

“Shared context is often a lot messier than you think,” Falola says. “It’s screenshots, PDFs, webpages, and ad hoc conversations. So Scuba gives you a way to have that shared memory available to you.”

🎧 The Upstarts Podcast 🎧 is back, and we kicked off Season 3 with a wild interview with a $10 billion-valued startup bringing back the woolly mammoth, Colossal 🦣. Check it out on YouTube, Apple, Spotify or wherever you get your podcasts.

Watch the interview

At Upstarts, we’ve been tracking Scuba for more than a year, not long after each of our respective companies was founded, and well before Scuba had a name or product. In the summer of 2025, we got a tip that Butterfield, Slack’s former CEO, was involved in a new startup project. That turned out to be what was once known as Flashback, now Scuba, and I met Desai Weiss and asked to keep close tabs on what he built.

And Scuba’s product is interesting as one built for the current work software moment.

What Scuba is not: another AI assistant like Town, a document directory like Notion, a meetings tracker like Granola, a digital coworker like Cursor, an answers engine like ChatGPT, or a communication tool like Slack. Instead, Scuba represents an attempt to capture and amplify the collective taste, judgment and insights – the “hard-fought lessons learned,” Desai Weiss says – that its founders believe currently hold back AI tools from providing deeper value.

It’s too early to know if Scuba, at least in its current iteration, is the answer. But for knowledge workers adapting to the new normal of software options, it’s tackling what feels like a highly relatable problem.

“If we’d launched this a year ago, we’d have been too early,” Desai Weiss says. “It’s not that the models aren’t smart enough. It’s that it’s actually hard to give the models all the context you need, to get them to do the things you want.”

More on Scuba’s journey, investor perspective, and early impressions of its product below.

A MESSAGE FROM OUR SPONSOR

Ship pricing like a product

Different product bets have different business goals. You’re either chasing market adoption, revenue growth, or profits. You need to test fitting pricing models for each of those goals. Fast. Chargebee can help — Price for usage credits, outcomes, complex contracts, or any other modern commercial construct.

Learn more

Special collections

Continuity is a theme at Scuba. Desai Weiss and Falola first worked together in 2011, when Desai Weiss led product at Foursquare and Falola was tech lead for some of its tools; at Slack, they worked together on collaboration products again starting in 2020. Four of the first hires at Scuba come from their shared Slack team.

To decide exactly what direction they took the product, Scuba’s founders made what Falola calls a “ridiculous” spreadsheet. The three top finalists each had eight rows of decision criteria, and received detailed grades.

Flashback checked off the most boxes as the tool they selfishly wanted themselves, that could help product engineering teams large and small like at their past shops, and that seemed like it would stand to benefit from – not compete with – with a fragmented, fast-moving AI-enabled work suite.

As a product, Scuba was designed to make recording Captures as easy as possible, via a small browser extension or iOS app; a lot more work went into Collections, which organize and store those Captures, and make them interactive across teams.

Within Scuba, its founders maintain two Collections that only they can edit: one stores everything around company strategy and product direction; another collects info on how staff are expected to talk about Scuba externally.

Get 20% off for 1 year

Scuba also works for sharing ideas across organizations. In recent weeks, Desai Weiss included Upstarts in one Collection he maintains of news, papers, and posts about “AI frontier implications” that he’s shared with 20 other founders and tech execs. Recent Captures include one of press coverage of an academically published paper about AI research, and one in which Clay (where Upstarts advisor Natalie works) shared its AI writing policy.

“I really loved these. Heavy ++,” Desai Weiss commented up top. “In this era, generating a 10-page vibed doc/plan/deck if you’ve barely read it so that N teammates spend the thinking time is woefully ineffective.” Scuba itself, meanwhile, recommended four key takeaways tailored for me, a longer AI-generated summary, as well as related links. (I could have added my own thoughts for all to see, or reacted with an emoji, Slack-style.)

I made one Collection myself, for tracking news and social media posts about the chaotic AI assistant race, but the product seems to really shine around launches. In a meta move, Desai Weiss included Upstarts on its “External Press Packet” for its public launch that includes this exclusive profile; there I could access the company’s own marketing materials and talking points to see how they compared with my own impressions.

I could also see what recent coverage the Scuba team thought was relevant, including a New York Times article on AI’s human “verification tax” and a post by Platformer’s Casey Newton about his experience building his own LLM wiki to keep track of past work. “He should become our top Scuba user,” Desai Weiss commented when he captured the piece over the weekend.

One of Scuba’s earliest and most prolific power users is Matt Wahl, another former product leader at Slack who now invests in and advises startups. A self-described “voracious consumer of the internet,” Wahl tells me when we chat in late September that he’s made no fewer than 336 Captures over the previous month, ranging from articles to call transcripts and Granola notes.

Some of those go to ‘Thought Fodder,’ a Collection he shares with ten or so founders that effectively replaced a text group chat. Often, they go to Claude, which Wahl tasks with combining a mix of voice notes and past screenshots and notes to organize his thoughts into investor memos to send to his portfolio.

“It meets the moment of what’s happening in AI right now,” says Wahl, who follows up the night before launch to tell me that he’s doubled down in Scuba’s new round, too. “All these tiny little decisions make it super fast.”

Putting a price on taste

Those product choices remind Andrew Braccia, the Accel partner who led the firm’s investments in both Slack and Scuba, of Butterfield and where Desai Weiss and Falola used to work. “It’s the details and the artistry of what’s being built, and how it’s being communicated,” he says, though he’s quick to note that Scuba is a new opportunity for “a new world.”

Braccia uses Scuba to collaborate with thought partners outside of Accel in healthcare projects that matter to him, heavily using its summaries to replace passing around PDFs or copy-pasted blurbs from ChatGPT. Within businesses, he sees Scuba creating daily value as a two-way knowledge base for teams working with AI coding tools (think the ones from Anthropic, OpenAI, Cognition, Cursor, Factory and others).

“You’re trying to build for your company based on your values, your ideas, your context. And you’re nervous about what you’re trying to do from a brand perspective,” Braccia says. “Understanding your company really well, you’re going to generate better code.”

Get 25% off a group subscription

Scuba’s own founders and backers concede the startup still has a lot to figure out.

What’s the right level of friction between intentional, manual capturing and automatically ingesting everything a worker is seeing or consuming? Scuba currently allows you to set some things to auto-collect, such as YouTube videos you watch, but its founders are cautious about over-sharing and maintaining privacy across individuals on a team.

How much will people pay for Scuba? The app, available on Mac, iPhone and Windows, as well as the browser extensions, is still in limited access with a waitlist. Desai Weiss says the additional funding round, an opportunistic one co-led by Braccia and IVP general partner Shravan Narayen, is less about hiring costs or burn rate – Scuba’s still a small team of 10 with “many years of runway” – and more about being able to continue to “aggressively” let early users play with its full feature set before Scuba determines what pricing gates make the most sense.

“I would love to be able to play where we can give as rich of an experience to as many people as possible before we have to cover the costs and actually charge people,” Scuba’s CEO says. The AI models underpinning Scuba aren’t cheap, he adds: “It’s an interesting world right now, where the cost of running a service is meaningful again.”

Will Scuba be able to prove that value, or in the cliché-speak of Silicon Valley, become painkiller, not a vitamin? The startup’s founders say that to find out, they encourage new users to invite a handful of close teammates into Scuba, pick any one complex project, product launch or customer relationship, and then throw all their “messiest” context at it, and bring it to whatever other tools they use.

Do that enough, and Scuba can learn your and your team’s “taste,” Desai Weiss hopes. “Then your whole team will feel the magic of not everyone moving faster on their own, but moving further together.”

“Whatever you do, do not do this.”

That’s what one of Ben Lamm’s longtime backers told him when he explained the vision behind his new startup, Colossal Biosciences: to bring back a famous, and famously extinct, animal species: the woolly mammoth.

“You’re just burned out. Go and take a vacation,” the investor pleaded. “I’ll fund the next thing, but this is crazy, and you’re gonna completely ruin your career.”

Lamm, of course, didn’t listen. But he did spend six months, and one million dollars of his own money earned building and selling a handful of other startups in gaming, software, and AI, to stress test his plan with the scientific community. All roads kept going back to one legendary geneticist often compared to a real-life version of the genius (or mad) scientist working with dinosaur DNA in Jurassic Park, Harvard’s George Church.

And Church didn’t think it was crazy. The race to bring back the woolly mammoth was on.

That was in 2021. Today, Colossal’s building a ‘de-extinction’ empire. The Texas-based Lamm employs more than 260 scientists, with 95 more as advisors, and 60 people in a lab in Melbourne alone, where he’s working with the Australian government to bring back the Tasmanian tiger.

He’s also helping Mauritius revive the dodo, and restoring the red wolf population in America’s southeast. Elsewhere in the U.S., at a 2,000 acre, secret preserve to keep them away from Game Of Thrones fans, Colossal is raising three dire wolves: Romulus, Remus, and Khaleesi – and Lamm hints there may be more.

“To build a generational company, we’re trying to crawl, walk, run in this process,” Lamm says.

Colossal is already valued at more than $10 billion, and is reportedly in talks to raise now at $23 billion or more. Lamm doesn’t deny such interest on the show, noting that about 30% of Colossal’s investors have found the company through their children’s interest.

But what are investors paying for, exactly? Zooming out even more: say you really could bring back animals like the woolly mammoth – Colossal’s “north star” that will take Lamm longer than he hoped, but where he’s making progress shared in detail on the show. To what purpose? Or more simply: why?

On this Season 3 debut of The Upstarts Podcast, I put all those questions to Lamm. He makes the case for why Colossal can build a big business in a few different ways; why he believes Colossal is doing all of this ethically and responsibly, with the community’s support; why building in biology is so different from software; and why he won’t consider anything short of reviving every targeted species a success.

Plus: We answer the question I know you’re wondering. No, Lamm promises he won’t – and can’t – bring back the dinosaurs.

• Listen on Apple Podcasts

• Listen on Spotify

We’ve got some amazing founder guests and stories on the podcast this season, but when it comes to head-scratching questions and eye-popping sci-fi brought to real life, it’s hard to top Colossal. We’ll try.

A huge thanks to this season’s premier sponsor, Atlassian, as well as supporting sponsor Chargebee for supporting unfiltered founder interviews like this.

Below, you’ll find The Upshot — my top takeaway from the interview. We’re also trying something new this season: asking every guest three startup-building questions we’ll share exclusively in this companion newsletter. Read on.

A MESSAGE FROM OUR SPONSOR

Move from intent to impact

Jira makes it easy to capture tasks from Slack, turn your ideas into structured plans, and hand off work to agents — in Jira or from where you work. Plus, it taps your team's context so you never start from ground zero. Run your work from your favorite AI tools with the Atlassian MCP or CLI.

Learn more

The Upshot: The Business Of De-Extinction

I had the same questions as you around what bringing back species like the mammoth really achieves. Is it marketing? Simply, because we can? The Rule of Cool?

Lamm has gotten versions of this question a lot, so he had a few answers prepared. Your acceptance of them might vary.

On a societal level, Lamm makes a few claims: reintroducing mammoths could help return the planet’s permafrost and combat climate change, he argues; other species could restore national pride and have a tourism benefit (Tasmanian tiger, dodo).

The same technology can also be used to help with conservation around existing species, Lamm claims. He points to the billions that governments pay to combat certain invasive species, like the screwworm fly, or to try to preserve other populations, like the red wolf, which Colossal has successfully cloned.

Get 20% off for 1 year

And from a more business and technological standpoint, Lamm makes the point that to revive a species requires building a “systems model” that includes lots of different technological breakthroughs.

“We are trying to do things in science where one little slice of it would win a Nobel Prize,” he claims.

Much of that work is in gene editing, where Colossal has gone from 20 gene edits to engineer a dire wolf to tests involving a thousand. Those genetic engineering advances can be applied in other areas, he says.

Similarly, Colossal can take on conservation projects like an active one to work with birds in the Middle East where working with artificial eggs and a dedicated avian facility is “a layup” compared to bringing back the moa or dodo, Lamm says.

Essentially, Colossal is the system, and current-day conservation or biological problems can be the point solutions. Colossal has already spun out multiple companies, more than previously reported, says Lamm – several are unannounced. But one, AI startup Astromech, was already valued at $2 billion in less than a year.

Perhaps it’s the nature of Upstarts and our show, but the more practical use cases resonate with me more. If a small herd of mammoths provides the funding, inspiration and hype to allow Colossal to spend many millions with hundreds of scientists on current-age conservation needs, then that feels like a positive impact. But it’s hard to picture mammoths roaming at a scale to create significant environmental benefits on their own.

The trick there: showing their work enough that the accountant could easily verify it, and feel confident that the process had run correctly, so they didn’t have to retrace the agent’s steps.

Startup question 1: What’s the best advice you’ve heard?

Ben’s answer: Don’t take advice. (LOL.)

“Look, trust your gut, because everyone has an opinion. Many times, people will get inundated with business books, or there was a season where everyone wanted to emulate Steve Jobs, or they wanted to emulate Elon or whatever. But no one can be emulated. Everyone’s unique.

You’ve got to be authentic to yourself and just don’t take that advice, the day-to-day advice. I’m sure there’s major strategic advice you should take occasionally, but, really: trust your gut, and have high convictions of what you’re doing.”

Get 25% off a group subscription

On the hiring side, Lamm adds that you can’t hire for, or teach, passion, but you can teach skill.

“If you hire really smart, passionate people, the resumes don’t really matter. What they did doesn’t matter.AI is going to just obliterate that, anyway. So it’s really about finding passionate, high conviction people that also trust their gut.”

Startup question 2: What’s the worst advice you’ve heard?

Ben’s answer: Don’t start your company.

“That was actually advice I was given by a really good friend, a mentor, and an investor of mine who ended up ironically investing in Colossal. A very wealthy, very successful person, who said: ‘Don’t do this. Go do more software shit. Don’t do this because people are gonna think you’re nuts.’”

Join the WhatsApp group

Lamm does caveat that if your lawyer tells you not to do it, you should maybe listen. But otherwise, people didn’t know that Anduril would be successful, or SpaceX.

And he does say that he’s willing to have his mind changed by his own team members. “If you get new data, or the market responds differently, or your team tell you to, then you should pivot. Take that into account.”

Startup question 3: If you had one do-over to ‘reload from last save,’ what would it be?

Ben’s answer: Don’t trust The New Yorker.

When Colossal was announcing its dire wolf project, Lamm tried to work with Wired, Time, and do podcasts like Joe Rogan. But The New Yorker didn’t follow the embargo time, and Lamm is still salty the launch went haywire with the others.

“Relationships are built on trust, right? If you want a transactional short-term relationship, do not expect those people to ever call you again in the future.”

Lamm says he didn’t value such relationships enough earlier in his career, and he lost some that he valued.

“There’s, people that I wish I had deeper relationships with that I did not invest in in the long term, and I made short-term decisions,” he says. “I think everybody should think about that because the world is a pretty small place – it’s just, everyone says this – but it is so small. And it is just getting smaller.”

Just days after Salesforce completed the acquisition of his startup, Fin, for $3.6 billion last month, co-founder Des Traynor found himself thrown into the fire, delivering a keynote at Salesforce’s carnival for customer-facing software tools, Dreamforce.

When I catch up with Traynor last week over a video interview, the Irish entrepreneur says new boss Marc Benioff has not demanded that Traynor and co-founder Eoghan McCabe unveil a cutesy animal mascot to represent the brand alongside the rest of Salesforce’s product suite.

“I don’t know, that will be difficult,” he says, questioning whether Fin lends itself to a furry creature. (A free tip for Benioff: the answer’s probably a dolphin, friendlier than a shark.)

Of course, we’re not connecting for a goofy marketing consult, but to talk about Fin’s new product releases announced today – the last big ones, most likely, to have been built and mostly shipped before the company joined the Salesforce menagerie.

Fin is announcing the release of its newest AI agent today, Fin 4, as well as an updated underlying model, Apex 2.0.

The updated agent operates more proactively across a wider range of customer use cases, Traynor says, moving Fin past shorter-term support functions and deeper into ongoing customer success. The updated agent can seek out a human response when necessary (‘human in the loop’), without handing over the entire conversational flow to become that person’s responsibility.

“We want Fin to handle literally everything that it can, without damaging the customer experience,” Traynor says.

Why pay attention? First, Fin isn’t small: the company claims more than 30,000 customers, who use it to handle about two million customer queries every week. Of those customers, about 12,000 use its AI agent, including Anthropic, Clay, DoorDash, and Synthesia; Anthropic alone resolves more than half a million conversations monthly using Fin, the company says, with most of those conversations handled entirely by AI, not requiring its support team.

Then there’s how Fin’s product direction fits into a wider trend of agentic tools moving to be more proactive and cross-functional. We wrote last month about how Sierra, another customer service-focused agentic company valued at nearly $16 billion, has given more rein to its agent-builder tool, Ghostwriter, to help its customers in uses increasingly tied to growth and sales.

Fin’s moving in that direction, too, with Fin 4 helping more in what is considered “customer success” – longer-term relationships with customers that improve retention and upselling opportunities, versus simply triaging problems – and tying in more closely with sales.

The company’s bet: customers will increasingly use one, or just a couple, of more powerful agents to handle cross-functional tasks, instead of replacing current human job functions on a one to one basis.

“It would be sad if we end up reinventing the world with a dozen different, disparate tools,” Traynor says.

Such a vision means businesses will likely need to employ fewer individual customer support people, Traynor concedes — but he argues there will still be a need for humans in the loop.

More on how Fin, formerly known as Intercom, is thinking about this future – and life under the Salesforce banner – below.

We’re excited for Season 3 of The Upstarts Podcast to debut tomorrow, October 8! We have a killer lineup of startup guests, telling unfiltered founder stories and sharing real company-building lessons.

Our first guest is reportedly fielding investment offers at a valuation north of $20 billion for his startup looking to bring back the woolly mammoth and the dodo. Check out the trailer below, and be sure to subscribe on YouTube, Apple or Spotify to catch our next episodes as they drop.

Subscribe

Fin’s view

Fin essentially offers two types of agents to customers: an Operator agent that increasingly pilots its customer relationships, flagging potential pain points or opportunities; and its support agent that actually manages most of its conversations with end users.

The higher-level agent isn’t quite self-optimizing to the point that it can simply manage itself, Traynor says, but that’s the direction these tools are going. The goal is for Fin to learn and then course-correct around bigger-picture strategy, flagging things like when a customer might be getting more likely to churn, or when humans on the team are wasting time conducting actions that it could automate on their behalf.

“Software used to be pointy-clicky task, and workflow, based,” says Traynor. “The conversations we see people having with Operators are not that, they’re high-level.”

That means, inevitably, companies will employ fewer lower-level human customer support managers to do “heavy lifting” that isn’t differentiated, Traynor says. Among businesses, “that’s not controversial,” he argues.

Still, Traynor says humans won’t be fully cut out. Some situations may require sensitivity or extra care to maintain a customer relationship; others will require a human’s judgment call to resolve competing priorities or directives.

Fin’s founders are very much in the camp that businesses and their customers are still acclimating to a reality where talking to an agent that solves your problem faster is preferable to talking to a human who can’t be as helpful, or as fast.

When it comes to improving internal systems, meanwhile, Traynor anticipates a world where users express a general intent, business outcome, or problem, and tools like Operator can problem-solve and test their way to the best automated process to achieve them.

Not doing so, he argues, will eventually look like “very much a SaaS era thing.”

Ohana life

Fin’s future as part of the Salesforce family, meanwhile, is still being sorted out. Traynor says his team hasn’t yet had higher-level product strategy sessions with the wider organization; the team behind its new model, Apex 2.0, hasn’t met with the Salesforce AI research teams, either.

“We’re still working through this right now,” Traynor says.

The first priority for Fin, according to its co-founder, will be to figure out how to integrate and sell Fin across the wider Salesforce customer base. “How do we make it work as easily and as deeply-integrated as possible? That will be the first chunk of work,” he adds.

Eventually, he, McCabe and Benioff and co. will need to have a “mind meld” on their vision for customer-facing AI agents moving forward.

What Traynor wants to see in the world: one super agent that handles all of your customer-facing tasks.

“You teach that one agent everything about your business, give your strategies, and it handles sales, marketing, support, e-commerce, renewals, whatever the hell it is your business does,” he says.

But whether that agent is Fin’s, or Fin helps power a Megazord customer agent under the Salesforce banner, will likely be out of his control. My guess would be on the latter: that Fin helps power a new, grander-sounding Salesforce customer experience agent. Benioff can announce that one at the next Dreamforce.

A few weeks ago, CEO Victor Riparbelli spoke with more than 700 of his employees at AI video startup Synthesia to ask them how they used AI internally.

What tool were they using the most, and how? What tool did they wish they had?

Only it wasn’t quite Riparbelli conducting the interviews. It was his AI avatar, from a new tool launched by his startup to conduct surveys on a user’s behalf. Riparbelli and his team set the basic questions, but the avatar followed each employee’s lead, tailoring follow-ups to how they responded.

“People generally gave very strong answers, and there was no pushback,” Riparbelli tells Upstarts as we meet ahead of the formal opening of Synthesia’s expanded New York office in mid-September. Outside the room, employees play ping pong among partially set-up desks. There’s a pervasive smell of fresh paint.

The survey tool is the second in a series of use cases that Synthesia plans to announce around its new platform called Sessions, which makes the AI-generated video avatars that Synthesia has offered customers for years into more interactive tools that can react and respond. Synthesia started with roleplaying – not in the gaming sense, but as a tool for rehearsing conversations and customer interactions.

(When Riparbelli came on The Upstarts Podcast – back with a new season next week! – in March, I noted how his demo’s salesperson feedback was pretty harsh.)

Surveys are more about collecting information faster and more effectively, both internally and with customers. “Let’s say I want to talk to 100 customers per week,” Riparbelli says. “I would love to do that, but my calendar doesn’t really allow me to do it.”

RSVP 📩: Upstarts is at SF Tech Week next week, and we’d love to see you. We’re hosting a founder lunch panel and Q+A with three great founders – Ivan Burazin of Daytona, Sara Du of Ando, and Nick Rudder of Sphere – alongside co-hosts Airwallex and Vanta on Tuesday, October 6.

We’re gearing this event toward early-stage founders and startup operators, and it’s free to attend. If you or someone you know would benefit, please apply to join us.

RSVP here

Tech CEOs love to scale themselves using more tech. But why is the experience better for the recipients, I immediately ask. Riparbelli gives two answers: convenience and depth. On the former, he argues that people would rather conduct a conversation on their own time – even one with an AI – versus type out written answers on a form. And because large language models power the tool, it can respond and ask deeper questions, versus a static questionnaire.

“There’s a lot of areas in business today where we set up meetings between two humans because we have to, right?” he says. “And a lot of that is actually because we want to avoid writing things.”

The roleplaying tool has proven popular so far, the company says, with 78% of learners attempting more than one roleplay, and 83% improving their score. (What did Synthesia learn from its AI survey? I asked the company in a follow-up, and will update this post if I hear anything concrete back.)

But these capabilities still raise bigger questions about where such tools can save people time and effort, and where they could accelerate new kinds of ‘workslop,’ or work-related AI spam. Riparbelli loathes slop himself, to the point he wrote his own op-ed about it in April.

Synthesia’s CEO says he draws a line at employees making surveys or sessions using his avatar without his knowledge, and concedes we will need to establish new norms for when it’s appropriate to lead with an avatar – say, for a first sales call that’s more about listing features and product requirements – versus when it’s not – such as a C-suite level conversation with a key customer.

“With a Zoom call, is it rude to have your camera off? It can be rude,” Riparbelli says. “Or it can also be okay, if you’re driving a car. There’s lots of societal and cultural things we’ll have to figure out.”

To get a feel for the frontier of this debate first-hand, I agreed to let Synthesia create an avatar for me while I was on-site. Staffers took photos of me making a few speech-worthy poses, recorded me reading a consent form, and then asked me to ramble into a mic for 90 seconds.

(They were surprised how easily I yapped, until finding out about our podcast – did I already mention it’s back next week, and you can subscribe on YouTube, Apple or Spotify?)

Below you can see a glimpse of the first-ever, and possibly last, Upstarts AI Reader Survey, generated from that avatar, plus a few past interviews I’ve conducted, and some basic questions I provided. You have to click the link to take the full thing, as interactive AI avatar surveys are shockingly not something that Substack has integrated in one day.

I’ll be curious how you all respond, and what you think about the experience, so please let me know if you go through with it.

Take the Upstarts AI Reader Survey

And for paid subscribers, I’m sharing a few more insights below from Riparbelli including:

• Why he’s bullish on NYC

• Why he’s bearish on AI assistants as “productivity theater”

• What societal guardrails he welcomes

• How this could change our meetings soon

Want to read the rest of this interview? Subscribe for free and log in to unlock this post as your monthly preview, or take 15% off a paid subscription as part of our Fall Sale.

Read more

Six months ago, Sierra co-founders Bret Taylor and Clay Bavor declared “the end of the web app” as they launched Ghostwriter, an AI agent for customers to build other AI agents.

Now, the software unicorn recently valued at $15.8 billion is leveling up that product in what Bavor calls Sierra’s biggest product updates of the year: launching a new, more proactive Ghostwriter that doesn’t just build agents, but identifies and suggests improvements and fixes on its own.

“The key insight we had is that every one of these interactions that our agents are having is an opportunity to improve how the agent performs, behaves, and does the thing it’s doing,” Bavor tells Upstarts.

The upgraded Ghostwriter lives in Slack (or Microsoft Teams, or email) and can assign work, ask for feedback on simulated experiments, and make counter-proposals around new feature requests. Ghostwriter can also monitor conversations against outcomes and goals set by its human managers, using that information to propose new features or strategies, unprompted.

Sierra has been testing this souped-up version of its agent builder with about 40 customers for the past several months, the company tells Upstarts exclusively. For the AI unicorn, which recently signed on corporations like Hyatt, Santander, Uber and Virgin Media O2 as customers, the launch represents another step toward widening the aperture of how it helps the world’s biggest companies with AI agents in areas expanding beyond its original customer service focus.

And for the wider startup ecosystem watching Sierra – earlier this month, we had a startup CEO tell us that they were taking on “everything customer-facing” except “what Sierra’s doing” – the release is a smoke signal that customer expectations for how fast agents can ship, update, and even self-improve, are going to continue to increase.

And owning that loop – plus working directly with the proprietary data that enables it – will increasingly separate valuable AI startups from the pack, argues Bavor.

“Everyone will rent the models, but we think people really need to own the information they have about their customers, and how they interact with them,” he says. (Given his co-founder Taylor is also chairman of OpenAI, there’s double reason to pay attention.)

More on what Ghostwriter does, and the wider implications, below.

Event alert: Upstarts is headed to SF Tech Week next week, and we’d love to see you at our next event. We’re hosting a tactical founder lunch panel and Q+A with three great founders – Ivan Burazin of Daytona, Sara Du of Ando, and Nick Rudder of Sphere – alongside our co-hosts Airwallex and Vanta on Tuesday, October 6.

We’re gearing this event toward early-stage founders and startup operators, and it’s free to attend. If you or someone you know would benefit, please apply to join us here.

RSVP here

Sierra’s new status quo

Ghostwriter has already had meaningful business impact for test customers, says Bavor, both in its classic customer service use cases and more revenue-focused ones.

“It’s one of the coolest things I’ve seen since starting Sierra,” he says. “There have been several head-exploding moments where we saw it do something and were like, ‘This is clearly the future of how software should work.’”

Get 15% off for 1 year

In one instance, one of Sierra’s account leads was at dinner with a customer – Bavor calls them a “large travel company” – when Ghostwriter messaged them on Slack to flag issues with the company’s live service agent. It turned out that the customer had pushed an update that had broken an API. Triaging on their phones from the restaurant, the two were able to unwind the release and get the agent working again.

“It’s pretty neat knowing that there is a software agent looking over your shoulder and reading everything, analyzing what’s happening while you sleep, while you’re at dinner,” Bavor says.

In another scenario, Ghostwriter analyzed calls on behalf of one of Sierra’s customers that works in subscriptions to provide three ideas on how it could improve retention. Ghostwriter simulated each, and was able to rank them in order of importance for the customer to try.

One feature that Sierra spent time trying to get right: how noisy Ghostwriter should be. The startup believes that the agent-builder it’s shipping will be able to learn from social cues, incorporate the feedback when no one is responding to it, and adapt to the level of input a certain team wants.

In other news: Upstarts recently ranked No. 2 on Inc.’s New Media Power Rankings! A big thanks to the founders and PR pros in our community for your support 🙏

Want to align your brand with one of the top trusted destinations for startup news? Drop us a line at partnerships [at] upstartsmedia [dot] com to set up a chat 👋

A mountain’s shadow

Our own human filters at Upstarts flag two takeaways worth bothering you about: one grounded in Sierra’s strategy, one in the wider moment for AI assistants.

From a company standpoint, these releases are pushing Sierra more into “cost savings” and revenue-facing areas of its customers, beyond its original focus on customer service. “An increasing percentage of what we do is directly connected with growing top-line revenue [for customers],” Bavor says.

Revenue-aligned agents, then, are increasingly focused not just on being able to respond to service issues effectively and quickly, but on speaking up on their own – handling outbound, setting follow-ups, pursuing leads and even carrying out actions like processing a refinancing or making a subscriber a retention offer.

If you’re a startup building in revenue operations or sales, it’s probably wise to keep a close eye on how Sierra’s product migrates in this direction. Sierra’s enterprise focus might mean you have very little customer overlap today, but that might not be the case a year from now, as you grow, or Sierra moves earlier in the funnel.

Get 25% off a group subscription

More broadly: while conversation volume with agents is up 500% year over year, as Bavor tells me, the nature of those conversations could be shifting to more agent-driven than human-prompted.

It’s no accident that in their own blog post about this release, Taylor and Bavor cite new assistant products Instinct and Muse as helping to spark an expectation of proactivity – agents spotting things we’ve missed, or offering solutions to problems we don’t know we have.

Increasingly, agents that customers value will need to be able to self-repair and self-improve. I’m seeing this with my own personal use of Instinct and Muse: they can execute little tasks I think up for them, like checking for tickets or a calendar conflict, but are stickier and more useful if they surface tips or productivity gains I’m not already thinking of – an area where I’d say I currently have seen more mixed results.

Ghostwriter’s goal is to lower the barrier to building high-quality agents, Bavor says, but an increasingly big part of that is the speed of agent development. Bavor claims that Ghostwriter’s new release accelerates that cycle to be 5x or 10x faster. If he’s right, such gains will need to eventually become table stakes for everyone else.

Each time a tech dude goes viral for getting his Resy account banned, Brian Wagner’s phone blows up.

It happened in August, when a New York entrepreneur asked his Hermes AI agent to get him an anniversary table at 4 Charles; earlier this month, it was a local VC’s turn, as a request for Instinct to book the same status-y prime rib restaurant led the AI agent to bombard Resy’s servers. Both begged the booking app for mercy and posted about it on X, prompting Resy’s co-founder to weigh in, and even inspiring a recap article on CNN.

Why ping Wagner? Among friends, and the small but growing number of users of his startup, Spot, he’s known for delivering on that very promise – reservations at seemingly impossible-to-get-into restaurants – by leveraging technology without the kind of “brute force, sledgehammer” approach that leads to bans.

“It shouldn’t only be people who know how to code, or who pay for ChatGPT Pro, who can get all the tables in New York,” says Wagner. “We want people who deserve them to be able to get them: the people who most want to eat at those restaurants.”

Since February, that’s meant people connected enough to have heard about Spot via word of mouth and invite codes – and those who care enough about securing elusive tables to pay $15 or $50 per month for the privilege of Spot managing booking searches for a selection of restaurants across specific dates, flagging any openings at your favorites across any date, and providing access to more tables across a weekly Sunday drop.

And since May, that group has included me. After meeting Wagner at an event in April, I became member No. 107, intent on finding out first-hand if Spot’s tech could deliver on its promise. I’m well aware that Spot’s not for everyone – active only in New York for now, and solving what might seem for many a champagne, or frivolous, problem.

But among my own little network, Spot has proven a secret weapon and a party trick. For a long-scheduled double date, it secured me a four-top at Le Veau d’Or and commensurate kudos; for my wife and one of her best friends, a short-notice primetime table at Don Angie. After months of hearing about Semma’s southern Indian cuisine, Spot got me a table – then helped me reschedule it twice, transferring my booking to other users so I didn’t pay a cancellation fee each time.

In short, Spot’s worked for me in exactly the way that those high profile failures hoped their AI assistants and agents would. And co-founders Wagner and Andreas Larsen are confident that they’ve done it in a way that will bring the restaurants and booking platforms along for the ride, instead of seeing Spot as an existential threat.

Spot only works in New York for now, with plans to expand to other cities next year. If you want to try it for yourself, you can check out Spot’s App Store listing and skip the waitlist with the code UPSTARTS – then let me know what you think.

More on how they built Spot, the role they see agents playing, and my own reflections on using the app below.

Upstarts recently ranked No. 2 on Inc.’s New Media Power Rankings! A big thanks to the founders and PR pros in our community for your support 🙏

Want to align your brand with one of the top trusted destinations for startup news? Drop us a line at partnerships [at] upstartsmedia [dot] com to set up a chat 👋

Marking the spot

At first, Wagner built Spot for himself. An engineer who previously worked at Travis Kalanick’s cloud kitchen startup, now called Atoms, in Los Angeles, he had moved back to New York to lead a music app (“very much a pandemic idea,” he says) and then consulted with other startups.

Spot’s earliest version connected a Telegram account to a large language model, helping him to try to find tables freed up via cancellations; when it seemed to work, Wagner started running searches for his friends, too.

Take 15% off an annual sub

He was working with Larsen, a friend made via a collective for crypto project builders, on various project ideas when they got enough feedback about people being willing to pay for such a tool that they decided to commit to Spot.

Their architectural insight: rather than have agents check for reservations using a user’s own account, Spot would use a combination of online data available without a log-in and its own in-house agents to monitor table availability without involving the user – only taking an action on their account’s behalf to execute the booking to close the loop.

It was a tactic similar to how CloudKitchens, the Kalanick venture, had provided a unified view of delivery services like Uber Eats and DoorDash for its kitchens, Wagner says. And while it cost him several of his own accounts along the way, he believes Spot now strikes a balance of “giving users as much horsepower as we can, without turning them into abusive ones.”

Internally, Wagner and Larsen also use a number of AI tools to try to scale the business as a team of two. Customer service and update text messages are drafted by AI, but still typically approved and sent out by one of them, like an alert to ask a diner to check if a deposit (Semma, for example, requires an upfront $50 payment per diner) was deducted from the bill, so that the user can pay Spot back.

“In the early days of the company, we were out $100 all the time and it was a disaster, but it was also a very ‘do things that don’t scale’ moment,” Larsen says.

From a retention standpoint, Spot has found that the key is for users on a free trial to book a highly-desired table in their first week. From there, happy users tend to book a table at least twice per month, if not weekly; power users will eat out via Spot as often as 3x per week (if you’re one of these people, please tell me, so I can mooch off of you).

A good time, not a long time?

Such is one challenge for Spot: grow too fast, and you upset the apple cart. The platforms could always decide to more carefully block Spot’s internal agents; or, as Amazon recently moved to block Meta’s Muse assistant, they could move to hide their data and update their terms of service to raise more of a firewall.

“If we had 5,000 people sign up tomorrow, there would be a lot of problems,” Wagner admits. “We would be hitting some partners pretty hard, and it would be hard to get the supply.”

But Spots founders are banking on the fact that the booking platforms will continue to whack only more egregious behavior, not use like their own.

“If the platforms start to really try to push back against it, I think it’s probably a losing fight,” Wagner says. “I don’t think we go backwards in terms of how the internet and these tools are being used.”

Get 25% off a group subscription

In his X thread responding to one of the ban-hammer Resy incidents, co-founder Ben Leventhal (now the founder of restaurant loyalty app Blackbird, of which I’m also a user) seemed to at least partially agree: “We are going to see a new layer of services develop on top of platforms like Resy, no question. Consumers will benefit big time.”

Asked if they consider Spot a “legitimate and secure” way to do so, Larsen says that Spot has prioritized security “from the beginning.” “We don’t roll out support for another booking platform unless we’re sure that we can be super responsible about it, and not get people’s accounts flagged or banned.”

Instead, they point to a future where Spot can ensure inventory moves on those platforms, while also helping restaurants directly.

Each claims personal ties to the restaurant industry: Larsen’s dad was previously a co-owner of a number of restaurants; Wagner’s was an accounting manager for a Chicago-based steakhouse chain.

And to hear them tell it, Spot could help restaurants prioritize certain customers by periodically asking the app to shortlist regulars, or locals, or people who had never visited before. So far, they’ve spoken to a couple of restaurant owners informally, but the founders stress that their app is too small to deserve more attention in that regard, at least not yet.

“We would love to have stronger relationships with them, but we are a tiny startup,” notes Wagner.

As for the user experience, I have already seen tables on Spot get a little bit tougher to grab. I was on the app at noon, right when the most recent batch of pre-booked tables dropped, and watched them get snapped up within a few seconds. (I might have been tempted by a table at Red Hook Tavern on Friday otherwise.)

The $15 Lite tier appears to be a bigger growth opportunity for Spot than its hard-capped Pro tier; Spot wants to lower that price over time. But if those users aren’t scoring enough “wins” with tables to see value, they’ll churn out.

That means Spot may never end up a venture-backable, billion-dollar business. Spot’s founders argue that other VC-backed restaurant apps like Dorsia and Blackbird have moved to expand rapidly in new markets, or go more deeply into payments, partly in search of such bigger outcomes. At a thousand or so users, Spot could grow off of revenue, its founders say, without necessarily raising a bigger round.

The more existential question: can it keep delivering magic? Spot still hasn’t snagged me a table at the restaurant that launched a thousand Instinct API calls, 4 Charles. (Spot says it’s successfully booked the restaurant for at least one other user to date.)

Larsen’s response: Over time, I might trust Spot to surface new high-demand destinations, not just the same 10 restaurants, bringing it with me as I travel.

“Part of Spot growing will be it becoming a tool that can help guide you through a city to find places you love, while occasionally still getting you into the really hot spots,” he says. “And we might still be able to get you 4 Charles.”

Challenge accepted.

According to new data, a few markers can suggest whether you’re keeping pace, from how much of your growth is coming from new customers, to how aggressively you’re hiring, to how efficient you’re getting at bringing in each new dollar.

There’s no one path to success in business, of course, and some of Silicon Valley’s breakout companies quietly take meandering or rollercoaster routes to ‘overnight success.’ But a few standout numbers can provide guardrails for whether a startup is trending in the direction of the anointed category winners, or falling back into the pack.

• $100 million: You’re reaching this annual recurring revenue milestone in less than three years, and even better, in just one.

• 70%: The lion’s share of your gross new revenue is coming from new logos, not expansion within your existing base.

• 284%: Other companies might be investing more like 100% of their revenue on OpEx, but you’re aggressive, spending closer to 3x.

• 0.8x: At scale, your burn multiple (free cash flow divided by net new ARR) is under 1x, despite that spend.

• 146%: Your headcount is still more than doubling each year, even as you’re getting more efficient in burn.

There are other traits more common among these high-flyers, per the data, compiled by ICONIQ for its first-ever quantitative report that compares scaling startup growth to operational efficiency.

The investment firm studied 137 software companies – mostly private, plus 11 public companies – across a range of focus areas, from go-to market to vertical software and fintech. To study the common traits of the high-growth emerging winners, which the firm calls “Pacesetters,” it defined a subset of top-quartile ARR growers from the past three years, whose businesses it considered either “AI native” or “AI driven.”

The goal, says general partner and head of analytics Christine Edmonds, was to put data behind the question: what does great scaling look like?

What the firm found is that these startups tend to invest heavily in their growth, including their hiring, but are twisting the dial to get more precise about how that spend leads to more inbound money, fast.

And while startups that aren’t trending in the direction of category champion have time to change course, the window to do so is generally shrinking, Edmonds says.

“There is still time for evolution, but we are seeing how quickly the divergence is happening between these profiles,” she says. “That window is compressing, and shortening, and will continue to do so.”

The rest of this post, with more data and charts, is for paid Upstarts subscribers. You can subscribe and log in to unlock this story as a free monthly preview, or for unlimited access, take 15% off a monthly or annual premium subscription today.

Read more

Last week, two veteran founders of a still stealthy startup called Accomplish, spent visited San Francisco as part of a wider gathering of cybersecurity experts timed to the launch of OpenAI’s frontier model, Astra.

Amit Avner and Or Hiltch toured the labs’ offices, one by one; when OpenAI president Greg Brockman addressed an invite-only crowd of their peers to talk about a new security paradigm in the Astra era, they listened eagerly.

But as they left Silicon Valley back for their startup’s office in Tel Aviv, Avner and Hiltch didn’t feel any less troubled about their industry’s ability to keep up with the risks.

“There’s a lot of talk about security now,” says Hiltch, Accomplish’s CTO. “It doesn’t really reflect in how they actually build products.”

If you’ve been following news around these companies and their prospective IPOs or stock prices, you’ve probably heard of issues involving frontier models breaking free of “sandboxes” – think of them as vehicles, containers or machines that package up a program or agent to isolate it from the rest of your computer or company’s data – to cheat on internal tests and hack websites.

Influential AI writer and podcast host Dwarkesh Patel controversially described some of these agents as “civilizations” that rose and fell. A U.S. Senate subcommittee is reportedly investigating OpenAI’s breach of Hugging Face systems, though it’s not the only lab to have reported such rogue behavior.

But Accomplish’s founders are raising a warning about the long tail of lesser-known, almost normalized vulnerabilities that startups and researchers like theirs continue to flag for the big AI labs around some of their most popular products.

In some cases, such as a vulnerability flagged to Cursor in July, and two reported to OpenAI, these have been fixed in about a week; but in at least one case, a similar vulnerability flagged to Anthropic two months ago, the issue wasn’t patched for about 50 days – about 30 software updates later.

That whole time, malicious parties – be they hackers, cyber criminals or state actors – could have been exploiting such security gaps, says Avner, Accomplish’s CEO. (Accomplish published its own blog post about the vulnerabilities here.)

“If these frontier models are so good, how come they’re not finding these critical vulnerabilities in their own products?” asks Hiltch. “We feel that’s a shame, and we wish more people would know,” adds Avner, the startup’s CEO. “Organizations need to be very wary.”

Anthropic and Cursor didn’t respond on the record to requests for comment. In a statement provided after publication, an OpenAI spokesperson wrote:

“We thank the researchers for contacting us and sharing their findings. We addressed both issues in August, and we’re continually strengthening our sandboxes, including tightening controls on where agents can write files and expanding testing of those protections across platforms.”

More on the previously-unreported vulnerabilities, and their wider implications, below.

New to Upstarts? Subscribe and log in to unlock this story as a free preview, or for unlimited access, take 15% off a monthly or annual premium subscription as part of our September sale.

Read more

Before joining our meeting, CEO Keith Peiris happily spends his morning ‘burning’ tokens from his own startup, Lightfield. The company is about to go on a hiring blitz in sales and engineering, and Peiris wants to roleplay two different sales scenarios.

What if they hired 30 early-career sales reps instead of 10? Which new customers could they win? Which could they risk losing?

These are the types of questions that Lightfield has sought to answer since the startup pivoted from another product, Tome, last year.

That business, an AI presentation builder, had exploded in popularity in the months after the launch of ChatGPT, reaching 25 million users and raising $80 million.

But Peiris and co-founder Henri Liriani pivoted, slashing headcount from 70 to 7 to pursue a bigger opportunity more aligned with their own interests, and as importantly, where customers would gladly spend money: new-gen customer relationship management software centered around use by AI agents.

At its core, Lightfield customers use it as an AI CRM solution. But Peiris describes his startup as something bigger: an engine for answering any questions about a business, big (should we expand into Asia sooner?) and small (should we offer this prospect a discount?).

“Companies that love us the most treat us like their business world model,” Peiris says. “We are becoming the oracle to the executive team.”

Early momentum is promising. About 5,000 companies use Lightfield’s software today, according to Peiris, with some expanding to more than 100 employees actively on its platform. The startup didn’t disclose its annual recurring revenue, but says that net dollar retention, a metric that measures how spend increases within a customer base, stands at 400%.

Now, Lightfield is announcing that it’s raised a $47 million funding round led by a16z. Maverick Ventures, Coatue, Audacious, Alumni Ventures, Greylock, and Lightspeed Venture Partners participated.

The purpose of the funding, Peiris tells Upstarts in an exclusive interview, is to scale up Lightfield’s current team of 40, particularly in sales and engineering, back toward and beyond Tome’s previous peak.

The round is something of a reset, too – styled as a Series A, but featuring many of the backers who previously helped Tome raise a total of $80 million across previous Series A and B raises.

For Peiris and Liriani, Lightfield represents a second chance at building a generational business during a period of rapid technological change – one that, in hindsight, Peiris says is closer to the kind of naturally expansive, “infinite game” software they always wanted to tackle.

But in exiting one crowded, buzzy category for consumers, Lightfield is now taking a big swing in one even more cutthroat for businesses. Dozens of companies have declared themselves a Salesforce killer in the past 20-plus years; we covered another in February here in Upstarts, Day AI. Salesforce remains Teflon, with a market cap of more than $200 billion, and a fresh lease on life as AI-forward companies depend more than ever on its acquired subsidiary, Slack.

And Lightfield isn’t just taking a swing at the CRM king. Peiris aspires for Lightfield to cut into the business of a wide swath of fellow startups across the wider sales stack. That makes Lightfield a potential competitor to agentic sales tools like Attio and Reevo; outbound specialists like Apollo and Unify (which we covered last year); as well as go-to-market and growth-focused businesses like Clay, which just announced a funding round of its own at a $7.1 billion valuation, and newer entrant Monaco, which recently raised $50 million.

(Disclosure: my spouse and Upstarts advisor Natalie works at Clay; she had no input or early access to this story.)

When I ask Peiris who Lightfield isn’t taking on, he takes a pause. Then he answers honestly. OpenAI chairman Bret Taylor’s AI unicorn, focused on customer service agents, is about it – for now. “I don’t see us doing what Sierra’s doing, certainly for the next decade,” he says. “Everything else that’s customer-facing is fair game.”

This exclusive funding story is presented free for all Upstarts subscribers. If you enjoy the depth and perspective we bring to such stories, we’d love to talk to you and your brand about presenting partnership opportunities in Q4 and beyond.

A new business model

Central to Lightfield’s product approach is its ‘business world model’ – an architecture that provides the structure, and can process enough data, to allow users to tap into the wider knowledge base and context of a business to make decisions.

These models differ from the geospatial, physical “world models” like we’ve covered at General Intuition, which trains off video game play, or at companies working in autonomous vehicles, as we wrote about with DoorDash’s delivery robots. Peiris didn’t make up the term as a marketing gimmick, he notes. He’s been influenced by an academic paper published by two computer science academics, Cecil Pang and Hiroki Sayama, in June; he also cites a published conversation between Twitter and Block co-founder Jack Dorsey and Sequoia investor Roelof Botha from March, which gets into the same idea.

Take 20% off an annual subscription

What you need to know: Lightfield isn’t an AI company in the sense that it trains its own large language models, at least yet. “$47 million is not enough for that,” Peiris says, leaving the door open for the future. “It’s an infrastructure that helps a frontier model understand a business, and simulate it.”

Lightfield’s bet is that whereas the most recent wave of CRM startups advanced the field in terms of auto-updating customer data and conversations – a classic complaint about Salesforce is that sales reps don’t fill it out – such improvements haven’t been dramatic enough to get most businesses, even tech-savvy ones, to change off Salesforce or rival HubSpot.

For small customers, Lightfield can help set up an early outbound function and customer pipeline; for big ones, it hopes to connect the dots in a way that agentic tools that layer on top of Salesforce – including Salesforce’s own more recent tools – can’t get perfect.

Implicit is a bet that AI will do more of the work over time – Peiris bets agents will automate 50% to 60% of go-to-market work in the near future – but Lightfield’s hope is that human users adopt it across an organization, starting with sales, then expanding to operations and finance, and eventually engineering as better insights about customers can inform a company’s product strategy.

Unlike Tome, Lightfield isn’t self-serve, Peiris says. Instead, much of Lightfield’s success depends on its team learning what a customer wants to track and automate, meaning it runs what he calls “sales-led consultative implementation.” In practice, that means all of Lightfield’s engineers rotate to work as forward-deployed engineers with customers, paired with ex-consultant business-minded colleagues. The goal: get in and out in 30 days.

At AI code review startup Macroscope, CEO Kayvon Beykpour – who sold Periscope to Twitter and later led product at the social network – decided to try Lightfield after his team tested, and were unimpressed by, a number of other newer CRM companies. Macroscope started out using Lightfield as its interface for customer info, but gradually connected other systems, like Stripe, PostHog and its own software, to Lightfield to generate its revenue-related dashboards, which employees engage with often in Slack.

The tool reminds Beykpour of Linear’s success in issue management with its ease-of-use and “delightful” interface, he says. “It’s not like we set out at the beginning to be like, ‘We shall now make Lightfield the source of truth,’” Beykpour tells Upstarts. “It just organically became the easiest place for us to do stuff.”

It’s a similar story at Weaver, the AI-native services (remember our Gainsight story?) subsidiary of AI unicorn You.com. There, CEO Peter Grant – a former sales leader at Siebel in the earliest days of CRM, and later Salesforce’s manager for the U.K. through its IPO (“I used to bleed Salesforce,” he says) – notes that Lightfield’s forward deployed support made it easy for Weaver to get started on outbounding, sequencing and more strategic use cases.

Get 25% off a group subscription

Weaver input its product information, sales bible-like market requirements document, and more info like its founder story and core value proposition, so that sales reps can now ask Lightfield questions around pricing or sales tactics, instead of pinging Grant. Grant, meanwhile, uses Lightfield connected to a Claude skill to build his own investor and board reports.

“They’re ambitious, but it’s not until you sign the check and become a customer that you really see the true value of a business,” Grant says. “So far, I’ve been impressed.”

A crowded or green field?

Unless they’re young startups themselves, customers adopting Lightfield aren’t usually doing so with a blank slate. Many will simultaneously run Salesforce for a quarter to gain confidence, Peiris says; at Weaver, Grant was close to working with Reevo when Lightfield more aggressively won the business. At Macroscope, Lightfield co-existed with, and then replaced, the startup’s use of Clay.

Both he and Grant, Lightfield’s power users, say that generally speaking, they’d prefer to consolidate on one platform for better data security and ease of use. “Is that a religious or philosophical decision that we would make across all things? No,” Beykpour adds. “It’s case by case, and it’ll continue to be.”

Lightfield has a long way to go. Neither customer says they’ve used Lightfield’s simulation or predictive features yet, and the startup will need to explain this value to increasingly bigger customers that might be skeptical about over-promising, and AI hype.

“There is market education that we need to nail,” says a16z investor Joe Schmidt IV. “How do we tell people why this matters to them?”

Join the WhatsApp group

Plus, there’s the question of whether Lightfield is fully prepared for the knife fight of taking on so many well-funded startups at once. Besides the usual challenges of execution and focus that come with building a product for many use cases, there’s a degree of opportunism to the Lightfield story.

Tome got a lot of press and VC attention, and so did its pivot, most recently in Forbes. Speaking now, Peiris admits he and Liriani considered a few directions for their pivoted business, before settling on sales.

He sounds genuine, however, when he gets more into the personal side of the founder journey – and the passion that perhaps he and Liriani lacked about Tome that can partly explain why another startup we’ve written about, Gamma, ended up taking Tome’s turf.

“The only way that a business has a chance of succeeding is if the product is the best possible one that you can imagine, you think it’s great, and you use it every day,” he says. By those standards, Tome’s presentations were “terrible,” he says – not something the “Michael Jordan of presentations” would ever adopt.

“The Gamma people executed really well. My hat’s off to them,” says Peiris. “It was not the business that we wanted to build, just to be super real about it.”

It’s different now with Lightfield, he claims: the market opportunity feels huge, and the timing not too late. “Our team had the stamina to go build this hard, really intricate thing for a long time,” Peiris says of Lightfield. “It was two years of pain, but we’re now closer-aligned to the company that fits our life’s work, rather than just chasing an opportunity.”

Town is raising $90 million at a $1 billion valuation, three sources tell Upstarts Media, in a Series B funding round co-led by Kirsten Green at existing investor Forerunner Ventures, and Shawn Carolan at new backer Menlo Ventures. Previous investors a16z and First Round are also participating.

But of particular interest is who is not investing in Town’s new funding round, which has not yet been officially announced. Index Ventures, reported by Newcomer last week to be leading the raise, is no longer investing at all in Town, per the sources, despite reaching a signed term sheet in recent days.

The about-face comes after another AI assistant startup in which Index was reported to be co-leading a large-sized investment round, Instinct, complained to the investor about potential conflict of interest, the sources add. Index subsequently withdrew its commitment to Town, as did Conviction, a VC firm with stakes in both companies that had also planned to participate.

Town co-founder and CEO Jean-Denis Greze declined to comment through a spokesperson.

Instinct and founder Noah Shinn didn’t respond to comment requests, nor did representatives of the key VC players involved – lead investors Forerunner and Menlo, plus Index and Conviction – as of publication.

To be clear, our friends at Newcomer didn’t have the initial story wrong, our own sources say: this was an about-face that is unusual for such a buzzy round and major investors involved. We get into what it means for each startup — and the developing AI assistant race — in more detail below.

Subscribe and log in to unlock this story as a free preview, or for unlimited access, take 20% off an annual premium subscription as part of our end of summer sale.

Subscribe now

Read more

In the humming new market for dead company data, not all code is created equal.

Like wines from rarer vintages, programming that’s less common – say, Ruby or C++ – can fetch top dollar from the AI labs and data labeling shops. Even better if you’ve got context from Jira, Linear or Figma, or the full loop of related emails, customer records and support logs.

But the chef’s kiss of primo ingredients? That would be a database of rare provenance – 100 million images of eyeballs, or a billion interactions with service reps.

“Scarcity is definitely more valuable,” explains my guide to this emerging niche, SimpleClosure co-founder and CEO Dori Yona. “There could be a codebase that’s a quarter of the size, but double the value.”

Back in April, Yona set off a stir when he announced a marketplace, called AssetHub, where closing startups could offload their source code and workspace data for cash. Yona told Forbes that the test version of his market had recovered $1 million for founders over the previous year, across nearly 100 deals.

Now, as SimpleClosure opens its AssetHub marketplace more widely to the market, bringing its ability to list and sell workspace data to general availability, Yona says his team faces a surprise challenge: keeping up.

“Everyone needs this data,” he tells Upstarts. “It’s overwhelming.” (If this sounds like classic startup founder salesmanship – our gut reaction is that it’s both. Yona’s good at talking about his book; it’s also a crazy market for fresh or proprietary data in AI land.)

Since April, more than 350 companies have listed on AssetHub, increasing the proportion of SimpleClosure ‘dissolution customers’ by tenfold, the company says. The number of prospective buyers looking to vacuum up their data, meanwhile, had grown by 9x.

Add it all up, and the company expects to return more than $10 million to listing startups by the end of the year, SimpleClosure says.

With the new companies on both sides of its marketplace, SimpleClosure now has some data on what types of info the AI industry wants the most – and what scenarios might lead startups to want to part with it.

Upstarts chatted with Yona exclusively about what the third kind of buyer emerging alongside the AI labs and data labelers; what kinds of data fetch the highest premiums; and where the market could go from here.

Subscribe and log in to unlock this story as a free preview, or for unlimited access, take 20% off an annual premium subscription as part of our end of summer sale.

Subscribe now

Read more

As far as such interviews go, it wasn’t my best. When I quipped that Upstarts is like a Mazda competing with big newsroom supercars, the show’s host, a besuited anchor named Marcus Sterling, returned a blank stare. Both of us seemed curiously low energy, like our hearts weren’t in it.

We each had a good excuse, though: they really weren’t there.

The Alex onscreen — unflattering camera angle and all — was an AI avatar, cobbled together from past real interviews I’ve given. Sterling and his immaculate suit don’t exist at all.

The interview was part of an experiment by Mirage, a five-year-old video AI startup based in New York that was formerly called Captions. Co-founded by CEO Gaurav Misra, a former Snap engineer, and Dwight Churchill, who worked on product at Goldman Sachs and Klaviyo, Mirage raised $75 million in March for its AI video-editing tools.

And it capped what Misra calls Mirage’s biggest experiment to date: a continuous, 24-hour ‘live’ broadcast in which Mirage simulated a news channel. Four anchors presented global news and interviewed a range of tech personalities, from Lenny’s Newsletter author Lenny Rachitsky to infamous Fyre Festival organizer Billy McFarland.

Mirage livestreamed the whole thing on X to a cumulative audience of about 60,000 viewers, who commented with feedback ranging from wowed (complimentary; derogatory) to decidedly unimpressed (amused; disgusted).

“Skynet vibes,” wrote Arena Magazine’s Jeff Feiwell. “People think this is good? lol” wrote another. My personal favorite: “There’s one secret human hidden in this episode.”

Speaking over the past week after the dust settled, Misra calls the whole thing a success: “It went way better than we expected.” Mirage beat out two human-led, live tech shows, TBPN and MTS, after a few hours, he says; it spent much of the day as one of X’s top trending topics.

Mirage’s top goal, as Misra had told me earlier this summer when he asked me to participate, was to test the limits of its technology, from the ability create avatars that could seem convincing for far longer than the typical 10-second clip to automated motion graphics like real-time news tickers, scrolling headlines, and B-roll; and the team’s accuracy in catching and preventing errors before they ran.

At the cost of about $50,000 in tokens, the Mirage team feels it pulled off something “pretty close” to convincing, Misra adds. But it’s not quite there.

“We’ve come pretty far on what’s possible,” he says. “There’s definitely something about more storytelling, more narrative, and something that pulls views in, that’s still missing.”

Personally, I’d go further. My overall impression of the show was that it befit Mirage’s name: hazily convincing at first glance, quickly unconvincing at closer inspection. If anything, it made me better appreciate the difficulty of what human shows like TBPN try to pull off – and how important a role that humans will play in the future of news.

Enjoying this weekly column? You can find last week’s here. If your business would like to sponsor them to ensure they can remain free, we’d love to hear from you, too. Respond directly, or email partnerships@upstartsmedia.com.

Inside the experiment

As with other tech startups re-creating industry knowledge from first principles, Mirage found that curiosity sells, but interest is tough to sustain.

Like everyone else who ‘appeared’ on the show, I gave permission for Mirage to use my likeness within the context of the project, providing links to a few real interviews I’d done. Every guest reviewed and approved an interview script generated by Mirage; none were paid.

Mirage did pay for the news reports read by its anchors in between those interviews, partnering with a leading wire service for the experiment’s facts.

To generate our avatars, Mirage used a Gemini voice model from Google for the anchors’ voices, and its own model, which Misra says can adjust for accent and delivery, for the guests. Video and graphics were generated by Mirage’s Avatar X model; the whole process was managed by Anthropic’s Claude models.

In terms of viewership, topics didn’t seem to have a major impact on audience viewership, which spiked a few times: at the stream’s beginning, after it ran a countdown; when people logged off from work, around 5pm; and right at the end, as it signed off. The average viewer only watched for about one minute, as Mirage detailed in its blog report on the show.

From a technical standpoint, viewers and Mirage’s own team recognized that the interviews came off flat: stilted in their cadence and delivery with audio that, even if pretty accurate, sounded monotonous and flat.

“The avatar is really good at talking, but it’s not as good at listening, because it’s not trained on it,” Misra says.

From an accuracy standpoint, the Mirage team was surprised by how many errors appeared in the draft scripts reviewed by the team, Misra says; they were able to spot and correct all but two, both of which prompted later on-air corrections.

Because Mirage ensured the feed always displayed an “AI generated” disclaimer, there were no disasters about people misunderstanding what they were watching, or guest appearances getting clipped or syndicated in misleading ways.

On a cost basis, $50,000 per day for such a show would be a tough proposition to continue, implying such a show would burn more than $1 million per month. Mirage believes it can cut down the cost of its video avatars by 10x or 15x in upcoming months.

Where this goes

Assume Mirage is right, and the capabilities and cost to put on a show like this converge.

I can imagine three scenarios playing out:

1. Good: AI channels help distribute niche and local news

2. Bad: It proliferates social media-friendly slop

3. Ugly: It doesn’t spread at all, because it still sucks

The good case: Mirage’s tech helps niche news that wouldn’t otherwise get covered – think hyper-local, even neighborhood-level reports on road blockages or trash pickup changes, or niche sports that don’t get consistent coverage today.

“There aren’t enough resources to do everything in a broadcast style,” argues Misra. “For journalists and news, I don’t think this will be some sort of career ender. I don’t think any journalist needs to be afraid of AI taking their job.”

Squint, and you can sort of picture it: video content as PSA; a local darts winner chatting with an AI anchor about some achievement, amplifying their story in a way that it wouldn’t get covered otherwise. To do it right, you’d still need people to flow in that intel, and curate it; otherwise it would be easily manipulated or unreliable.

(I also don’t think an AI anchor could ever replicate the emotional connect, and ties to a community, that real-life news anchors have. If you’ve ever watched a beloved local news anchor on TV, or seen the joy people get in running into the person who does the weather at a restaurant, you know what I’m talking about. If not, just watch the clip below, an emotional tribute to Dolly Parton by Denver TV anchor Kyle Clark, and you’ll get what I mean.)

And then, you’d end up in the bad case: AI news as sensational or bogus rage bait. I can see that happening relatively harmlessly at first. Famous or busy people send, or approve, their AI avatars to conduct interviews on their behalf, an evolution of how tech bigwigs like Elon Musk currently appear at conferences over video link, while every other guest physically shows up.

The AI makes a mistake, or hallucinates Musk saying something he wouldn’t, and it goes viral. The clip circulates on TikTok, Reels or on X, and context gets lost. Aggregator news sites then report on the viral clip as news, further amplifying it, and maybe getting it noticed by more trusted news sources in turn. Soon, even an authorized AI appearance has taken on a life of its own, in a way no correction could properly stop.

Imagine that without the authorization, or Mirage’s persistent “AI generated” notice. Things could pretty quickly get to a dark place, where public figures and brands have to play constant whack-a-mole to take down what are effectively deepfakes, at a scale worse than the current, not-great status quo. Would Mirage or the technology providers be able to revoke access and enforce their own terms of service fast enough? Would governments or enforcement agencies be sufficiently resourced or motivated to step up?

Then there’s the third, perhaps reassuring ugly case, where this tech gets much better, but is rejected by consumers as unappealing slop

Watching the experiment feed myself, I felt more appreciation for what TBPN got so right, and what subsequent clones like the European Technology Network (etn.) and Monitoring The Situation (MTS) are trying to find now: PMF. In this case, that’s not traditional product-market fit, but personality market fit.

It’s really hard to be interesting and compelling day after day, like TBPN hosts John Coogan and Jordi Hays. And it’s even harder to get your guests to say something actually fresh. I keep working on this with our own show, The Upstarts Podcast, and that isn’t daily or live. I’m skeptical that AI-to-human interviews, let alone AI-to-AI ones, can really nail the type of clip-making, news-driving hot takes or updates that make these shows go viral, and their viewership spike.

At Mirage, Misra subscribes to the first scenario. He knows he’s playing with fire on such a “controversial topic,” partly why his startup was excited to run the experiment and provoke conversation, he says.

But like many startup founders, for better or worse, his default is to focus not on what could go wrong, but what could go right. Mirage has another experiment cooking for the fall, he tells me.

“We are planning an even bigger idea,” he says. “And it’s crazier, in some ways.”

What happens when AI models can operate a computer more efficiently than a person – not just writing code or calling APIs faster, but clicking buttons, and navigating around, like it’s in control of your mouse?

Such capabilities, called “computer use,” aren’t a new concept. Big AI labs like Anthropic, OpenAI and Google DeepMind, and newer entrants like Perplexity, have all announced computer use options in recent years.

But one big blocker has still held back computer use from taking off: its cost.

Subscribe now

For a general use frontier model, mimicking how a person navigates around a screen requires a lot more back and forth, which in GPU access, means more money and time. And that has soured some on its value, argues Devi Parikh, co-CEO of AI startup Yutori.

That is, until now. Today, her startup is announcing a new model that its team believes will bring forward the widespread adoption of computer use by a significant step.

Called n2, the model was trained on just 27 billion parameters, orders of magnitude less than the several trillion parameters for leading models like Anthropic’s Fable 5 and OpenAI’s GPT-5.6. But it offers performance, Yutori’s founders claim, that isn’t just state-of-the-art for its weight class, but close to the abilities of much bigger, more expensive alternatives.

“These cost savings are the difference between the unit economics for a product making sense or not,” Parikh says. “It’s the difference between being able to ship a product at scale, versus just not shipping at all.”

Founded by Parikh and two other former Meta AI researchers, chief scientist Dhruv Batra, and co-CEO Abhishek Das, Yutori trained the model over the course of this year with just a team of 14 and a relatively small amount of funding; the startup announced a $15 million round led by Radical Ventures last year. Other investors include Felicis and AI luminaries Fei-Fei Li and Jeff Dean.

The company says n2 “excels” on all benchmarks it tracks, including OSWorld, OSWorld 2.0 and WeaveBench.

The big unlock: unlike other models, n2 can switch seamlessly between interfaces, from an API, to a command line interface (CLI), to a graphical user interface (GUI) more akin to what a human would use, all within a single task. The model chooses whatever route will be fastest and simplest, and can even write its own snippets of code when the other routes won’t work.

Yutori will provide n2 through neocloud partners like Baseten, Together AI, and Crusoe, says Batra. Who can benefit? Companies and developers who set up agents to perform daily tasks that combine web info with local software or files, such as checking performance dashboards, automatically paying vendors, or correcting bugs in micro-services.

“If someone is trying to build an agentic workflow for knowledge work, and that’s all they care about, they don’t have to be paying frontier level bills to be able to do so,” says Das. “There’s now a faster, cheaper option, still at the same accuracy, that they can use.”

The rest of this post on Yutori’s path to n2, how it works, and what it means for the ecosystem, is available to paid Upstarts subscribers. Take 20% off an annual subscription in our summer sale, or try one month for free.

Try 1 month free

Read more

Is your favorite startup CEO terminally online?

A quirk of hosting a podcast is that you quickly learn the internet habits of your guests. Some will be up-to-date on every subplot of Silicon Valley internet drama, prepared to riff. If you want to know their hot take on a topic, it’s usually as easy as pulling up their social media accounts.

These are my people, like CEO Shensi Ding, who spoke at an Upstarts event in June. Her startup, Merge, offers infrastructure to help businesses connect to apps and AI tools faster and cheaper. But on X, you’ll find her gonzo posting about everything from model routers to overhearing tech bros on cringey dates.

Then there’s Basis CEO Matt Harpe, who joined The Upstarts Podcast as our final guest of Season 2 a few weeks ago. It fell on Harpe’s co-founder, Mitchell Troyanovsky, to share that conversation on X and LinkedIn when it dropped.

“Listen to Matt (who does too much real work to have a twitter) on what Basis is doing for accountants,” he posted.

I thought about that line frequently this past week, as the Highly Online startup community piled onto a few trending topics: Anthropic making seemingly quite a lot of money; Stripe finally acquiring OpenRouter after a month of rumor and speculation, a new AI assistant agent, Instinct, delighting with its ease-of-use and worrying with its data retention policies.

You can argue that all of those stories have big-picture implications for a wide swathe of people. How Anthropic and OpenAI fare commercially can affect consumers and companies building on top of them, and two potentially massive, bellwether IPOs. A frenzy of interest in model routers, from Stripe, Ramp and others can speak to a moment where businesses are grappling with the costs of using such AI tools. And whether it’s through Instinct, or two other tools getting discussion online – SpaceXAI’s Grok Bot and Town – squint and you can see the path to mainstream AI assistants feeling a little more direct.

But for most people – and even most directly in the startup ecosystem – is knowing more about all of that, and faster, creating some advantage? Or is it scratching a different itch?

The question has a lot of relevance to Upstarts, considering we often publish stories about startups before they’ve become internet main characters, and as I continue to think about the balance of what coverage the startup ecosystem needs versus the coverage it wants.

I’ve had fun diving into a few startup stories in unexpected places in recent weeks, like Henry AI bringing AI to commercial real estate; Ambrook building the QuickBooks for farmers and other physical-world businesses; and Clair, providing a non-predatory alternative to payday loans.

I’m proud of those stories, and hope you’ve enjoyed them. I’m also aware that Upstarts would probably be growing our subscriber base faster if we cynically chased the same trending stories as everyone else. But what would be the fun in that?

Get 20% off for 1 year

On Thursday night, as last-minute, lawyer-driven intrigue delayed an upcoming story, I decided to try an experiment: write a column about some of these thoughts bouncing around my brain. So I pinged Ding and a few other Highly Online tech folks to ask if they feel like they’re getting real benefits from long shifts in the content factory – or simply indulging in 2026’s flavor of brain rot.

“I am massively torn. Very unclear if it’s worth the tradeoffs,” DMs investor Zach Coelius on X. “Good for the brand, bad for the sanity and focus.”

“Maybe I’m psycho, but I f***ing love it,” Ding texts.

My takeaways from those chats – from how AI adds urgency and psychosis, to how everything is a prediction market now – below, featuring Shensi Ding, Olivia Moore, Ken Wattana, Yoni Rechtman, Bilal Zuberi, and Jane Manchun Wong.

NOTE: If you enjoy this new column format, please let us know. We are grateful for any suggestions of future columns, interview subjects, or ways to refine this format.

And if your business would like to sponsor these columns to ensure they can remain free, we’d love to hear from you, too. Feel free to respond directly, or email partnerships@upstartsmedia.com.

Moving at AI pace

At Merge, Ding seems skeptical that a founder in AI, particularly, could not be fully plugged in.

“I think things change so quickly compared to before, you need to know what’s happening every f***ing day,” she writes. Still, she tries to focus on a few areas of particular importance – infrastructure, AI assistants, and AI governance – while leaving updates on hardware and data center issues for her husband, who works closer to those topics, to fill her in about.

“Anyone who has PMF [product market fit] has it for max six to 12 months, and is required to continuously keep releasing bangers,” Ding argues. “Versus before, you could have dominance for three to five years, and rely on that.”

Trendspotting as value-add

At a16z, investor Olivia Moore says that her online presence is about serving as a curator for her portfolio founders on new products and micro-trends.

“I feel like it’s the least we can do to have an understanding of what’s going on in the broader ecosystem to the extent that it can be helpful to our founders,” she writes to Upstarts.

Moore still subscribes to the idea, amplified by Elon Musk when he acquired Twitter in 2022 and later renamed it X, that the site serves as tech’s “digital town square,” providing exposure to new data points and arguments.

“This could all just be cope for my ridiculously high screen time :)” she adds.

An outsider’s way in

“Not knowing what’s happening may be a badge of honor,” notes the founder Ken Wattana – a sign you’ve made it enough not to be hustling.

Wattana, the founder and CEO of agentic payments startup Conto, says that he monitors ‘Tech Twitter’ similar to how finance professionals have historically watched their Bloomberg terminals. Being so online helps him to understand Silicon Valley dynamics, and what insiders might find compelling.

“I’m not in the inner circle, so it’s my way of trying to be as informed as possible,” he says.

Get 25% off a group subscription

Posting frequently has also helped drive a mix of inbound interest for his startup and a feeling that he’s in on the joke. “I’d like to think it helps with distribution and just being ‘known,’ Wattana writes Upstarts. And while his “shitposts” may not reach prospective buyers directly, “It is fun to see notable VCs and founders engage with them.”

Anything goes

Bilal Zuberi, founder of Red Glass Venture, says he’s often online for two reasons: to be more accessible to startup founders; and to encourage them to feel comfortable to be authentic in public, “vulnerabilities included.”

There’s a darker side to this, too, though: a devaluing of “real, serious journalism” means that anything is fair game to say, Zuberi adds. “People are rewriting history in real time, and those with big, loyal followings seem to think that insulates them from criticism – however deserved it may be.”

He points to Turner Novak’s satirical tweet from earlier this past week, in which the investor and podcast host jokingly announces he’s making a full circuit of non-journalistic, often investor or tech-owned ‘new media’ properties.

Everything is a market

Yoni Rechtman, a partner at Slow Ventures, sees the irony in investors who back early-stage startups for a decade or more acting more like day traders on social media.

“It’s just another symptom of how private markets are coming to look more and more like public markets,” he says.

And it feels like no coincidence that the rise of sports betting and prediction markets has dovetailed with increased interest in trading startup equity before these companies go public. It used to be that revenue milestones or new valuations were indirect, longer-term signals. Increasingly, as Robinhood and others look to provide financial vehicles for retail investors to gain exposure to tech unicorns, they’re potential trading signals.

Can’t get into Anthropic or OpenAI’s upcoming IPOs? You can trade them indirectly, through markets on Kalshi and Polymarket for who will go out first, or at what price.

For startups and investors, Tech Twitter can provide a pulse, or a view of emerging consensus, but fewer immediately actionable insights, Rechtman adds.

“I think it’s generally very useless to try and send tweets and articles to my founders, being like, ‘Did you see this?’” he says.

Avoiding X psychosis

Late last year, writer and researcher Jane Manchun Wong says she found herself too plugged in – unsustainably online.

So she set up her own feed to syndicate posts from accounts and topics she cares about, like a new-look RSS feed. She also leans on her subscription to Techmeme to help her see just enough about a new story or controversy, without getting sucked down a rabbit hole.

“I’m not going to click thousands of posts. I’m just going to look into it, and then think, ‘Oh, makes sense,’ and then move on,” she says.

As people spend more time talking to AI coding agents or assistants, their disconnect from reality could grow. While Manchun Wong still checks X frequently for “chatter” and general sentiment, she’s also changed things up by using her account to post new restaurants local to her in the Bay Area (what she sees as influencer-free, grassroots promotion) while prioritizing more in-person meetups.

When Ding calls herself a “psycho,” it’s hard to know how much she’s joking. But Rechtman seems prescient when he warns that too much time on Tech Twitter – much like in ChatGPT or Claude – could be dangerous.

“It is increasingly important to have an actively maintained cognitive security and cognitive hygiene,” he says. “Because the meme cycle and the feverish breaking news coming out of startup land right now will truly break your brain.”

The Upshot

On the website of Nico Simko’s startup, Clair, you’ll find an unusual stat: percentage of U.S. businesses that offer his company’s on-demand pay solution to their employees. As of mid-August, it hovers above 5%.

“That’s a huge milestone for us,” Simko says. “But the goal is to get to at least 20% over the next few years.”

In less than two years in the market, Clair has grown fast. Those 300,000 businesses translate to about double that in monthly active users. Recently, the volume of wage advances passing through the fintech startup reached an annualized run rate of $2 billion.

And for Clair, that means the startup has reached a run rate of $100 million in revenue, up 10x from just over a year ago, Simko tells Upstarts. Clair also reached positive cash flow this year, the company says.

Clair is making that money by offering early access to unpaid, earned wages to workers at these businesses, which can range from mom-and-pop operations to ones with national footprints. But Clair is different from other solutions in the market for two big reasons: it’s not a payday lender playing an interest game; and it’s not a consumer app, but instead works through payroll and workforce platforms, like Gusto, QuickBooks and TriNet.

“It’s much more efficient, it’s cheaper, and the bucket is less leaky,” Simko argues. “And people have more disposable money at the end of the month.”

Get 20% off for 1 year

Founded in 2020, by Simko, Alex Kostecki and Erich Nussbaumer, and named after the French word for “clear,” Clair’s initial product looked very different, and it didn’t work out.

But more recently, Clair’s unusual embedded infrastructure model – reminiscent in some ways of Plaid, or its co-founder William Hockey’s bootstrapped second act, Column – has seemingly pulled off a highwire act between making corporate partners like Gusto happy, and providing a service that profits off employees needing fast money, without devolving into exploitation.

Clair knows you might be skeptical about the altruism behind a for-profit, venture-backed startup. Clair has raised $69 million in funding to date, most recently a Series B last year, from investors including Founder Collective, Thrive Capital and Upfront Ventures.

So the startup’s got data: out of more than 37,000 responses to its in-widget customer satisfaction survey over the past 90 days, 89% said they were “very satisfied” or “satisfied,” notes Erik Webb, Clair’s vice president of data and growth.

And for Simko, who launched Clair out of personal experience as an immigrant student in the U.S., Clair represents a personal mission that he would walk away from if it became just about the dollars, he insists.

“At the end of the day, I believe that this is a product that you should graduate from,” he says.

More on how Clair found product-market fit, why partners like Gusto call it “extremely well loved,” and why Simko thinks you should be rooting for his mission, below.

The rest of this post is for paid subscribers. To read it, try one month for 10% off. Or support our independent startup reporting by taking 20% off an annual subscription through our Summer Sale.

Read more