☕ From a demo stage to a $1.8 trillion debut

On Monday, 552 people signed up to spend a workday in San Francisco watching 20 founders demo their startups live, Wi-Fi gods willing.

By Thursday, Anthropic was trading on the public market at a value of about $1.8 trillion.

That's the whole AI economy in one week: from scrappy demos in a crowded room to one of the biggest debuts Wall Street has ever seen. And if you looked closely, both ends were telling the same story about where the next wave of value is going.

Here's the short version:

  • The demo stage had a surprise theme. Hardly anyone was showing off a shiny new agent. Most were building the plumbing that makes agents safe to use at work, and we think that's one of the most interesting places to invest right now.

  • Wall Street said "yes" to AI. Anthropic's debut, plus Nvidia doubling its revenue in the middle of a leadership shake-up, says demand isn't slowing down.

  • Finance is where agents are getting paid. Two startups building agents for finance teams raised $530 million between them this week.

🎤 Super Demo Day: what we learned

20 startups. Four demo sets. 552 registrations. Presented by Open Source for AI, with Gravitational Ventures and AWS. Huge thanks to every founder who demoed live.

The surprise: almost nobody pitched a flashy new agent. Most were building the plumbing that makes agents safe to run at work. Two camps:

1. "Can I trust this agent?"

  • Langfuse: shows what your AI app is actually doing

  • SuperBryn: catches voice agents failing on accents and background noise

  • Guild AI: controls what agents can access and tracks what they cost

  • Descope: identity and sign-in

  • Ipsion AI: keeps a human in charge of an agent team

2. "Can this agent reach my data?"

  • Apify: pulls data from the web

  • CData: connects agents to business systems

  • PuppyGraph: queries existing data as a graph

  • Debreu: turns dealmakers' emails and calls into usable data

Why it matters for investors: this plumbing works with any AI model, grows as customers add agents, and is hard to rip out once installed.

📈 Meanwhile, on Wall Street

While founders were demoing in San Francisco, the public markets were having a big week of their own.

Anthropic went public on Thursday with a market cap of about $1.8 trillion. Back in Issue #2 we said public investors were about to put a price on the AI boom. Now they have.

Nvidia kept rolling. In its first earnings report since Jensen Huang resigned, the company posted $115.8 billion in revenue, up 103% from a year ago, and expects around $130 billion next quarter. Interim CEO Colette Kress is running the company while the board searches for a permanent replacement, and a new board committee will oversee national security and trade compliance.

So what does this mean for startup investors? Three things. Big public debuts tend to free up cash for venture investors, and some of it finds its way back into young companies. Strong public valuations also make it easier to justify prices for AI infrastructure startups. But the bar goes up too: when anyone can buy shares in the AI giants directly, a private startup needs a sharper answer to "why you, and why now?"

🚀 Startup Spotlight: Staris AI

Here's a question every engineering team is quietly asking: if AI is helping us write code twice as fast, who's checking it twice as fast?

Today's answer usually isn't great. Security scanners spit out long lists of "possible" problems that teams spend weeks sorting through, and manual security tests happen a few times a year at best.

Staris AI takes a different approach. It behaves like an attacker against your running application, actually tries to break in, and only reports problems it can prove are real. Every finding comes with a ready-to-merge fix, and the company says it can deliver results in as little as 15 minutes. It also says it doesn't train its models on customer code, which matters a lot to security buyers.

Who's behind it. Founder and CEO Adam Cecchetti and the leadership team have worked at Amazon, AWS and JPMorgan Chase, and at security companies later acquired by Accenture and GitLab. Staris raised a $5.7 million seed round led by Freestyle VC in January 2025.

Why now. AI is helping people ship more software, faster, and a lot of it is never reviewed by a security expert. That gap is exactly where Staris lives.

📡 Follow the money

A few rounds from this week worth knowing about:

  • Model ML raised $250M at a $1.5B valuation for agents that handle financial modeling and due diligence. Investors are paying up for agents that do expensive professional work.

  • Payhawk raised a $280M Series C at $2B to build a finance operating system run by AI agents. The finance back office keeps showing up as a favorite place for agents.

  • Pretxt came out of stealth with a $5M seed to supervise AI agents at work in real time. Early money is flowing to agent oversight, the same theme we saw at Super Demo Day.

🔢 One number: $530M

That's what Model ML and Payhawk raised between them this week, both for agents that do financial work. Why finance? It's one of the easiest places to measure what an agent saves, and buyers there have big budgets.

🧭 Where we're looking next

If you're looking for your next company to dig into, here are the four areas we're watching most closely, and the sign we'd look for in each:

  1. Control centers for agents. Tools that manage what many agents can access and what they cost. We'd look for big companies standardizing on one platform.

  2. Testing and monitoring, especially for voice agents. Voice agents trip over accents, noise and interruptions in ways text agents don't. We'd look for pilots turning into real contracts.

  3. Connecting agents to company data. We'd look for adoption through open standards like MCP.

  4. Agents for finance teams. We'd look for customers expanding from one workflow to several.

What could change our mind: if the big cloud providers or AI labs start giving these tools away for free, the opportunity for standalone startups gets a lot smaller. Worth keeping an eye on.

💬 Your Turn

For investors: want to dig deeper into any company or area in this issue, or hear more about our portfolio? Drop a note to [email protected]. We'd love to hear from you.

For founders: building in one of the areas above? Hit reply and tell us about it. You might end up in a future issue, or on stage at our next demo day.

Know someone who'd enjoy this? Send it their way.

Keep Reading