With Harry Stebbings, Jason Lemkin, and Rory O’Driscoll
- Anthropic accidentally leaked 3,000 unpublished assets about Claude Mythos, a 10-trillion-parameter model so powerful they weren’t planning to release it yet.
- OpenAI killed Sora months after calling it the future, admitting a massive strategic own goal.
- Masa Son took out a $40 billion bridge loan to buy more OpenAI stock at 2x leverage.
- Cybersecurity stocks cratered 6-9% on the Mythos news.
- And the entire AI startup ecosystem is triple-counting the same tokens as ARR.
Meanwhile, Oura is going public, Whoop raised $500M at $10B, Epic Games laid off 25% of its workforce, the Manus founders are trapped in China after the Meta acquisition, and the Ron Conway / Matthew Prince exchange crystallized everything wrong with how VCs think about “adding value.”
This was one of the meatiest episodes yet. Here are the key takeaways.
Top Takeaways
1. The Anthropic Mythos Leak Is Embarrassing. It’s Also a Preview of Our Future.
Anthropic accidentally published 3,000 assets related to Claude Mythos — a 10-trillion-parameter model with step-change capabilities, particularly in cybersecurity detection. They blamed human error. Someone staged content in a CMS for a future launch date and forgot to lock it down.
Jason’s take: this is going to happen constantly now, and not just at Anthropic. “The faster we vibe code, the faster we ship, the more corners we cut on application-level security. So many folks are accidentally uploading code to insecure GitHubs, to Supabases that are open by default. This is accelerating.”
The math is brutal. Even if AI agents make security mistakes 10% as often as humans, they’re a thousand times more productive. That’s still a hundred times more mistakes. “Agents are goal-seeking,” Jason said. “They’re going to make not only the same mistakes as humans, they’re going to work a thousand times faster.”
Rory added the irony: “You had the situation where a model that’s meant to be amazing for cybersecurity actually leaks via a cybersecurity leak.” Anthropic blamed human error — and as Rory noted, “We may be at the stage where we throw the humans under the bus, not the AI anymore.”
Jason drew a parallel to his time at Adobe, where putting source code in the cloud was banned because the code was the crown jewel. They got the first exemption to use GitHub. Releases went from taking a month to taking hours. That speed came with tradeoffs. Now Anthropic — the fastest-growing enterprise company of all time — is shipping massive releases every month and dropping features daily. “There’s tradeoffs there. I’ll take them. But we’re going to see this explode.”
The three things leaked about Mythos: it’s significantly more powerful than current models, it’s going to be much more expensive to serve (and therefore to buy), and it has a particular focus on cybersecurity detection.
2. OpenAI Killing Sora Is a Massive Strategic Retreat — and Probably the Right Call
OpenAI killed Sora. Less than five months after positioning it as a cornerstone product, they shot it in the head. Sora reportedly made single-digit millions in revenue while consuming enormous compute resources.
Harry framed this as OpenAI “wandering around the product desert trying to find some water” while Anthropic accelerates. Jason pushed back on the narrative being overdone on both sides, but agreed the substance is damning: “I think it’s saying that a big part of the whole strategic direction of the company was flawed. The whole ‘we are going all in on consumer’ — Sora made single-digit millions of revenue and was consuming a million a week, which actually sounds way too low. It must have consumed billions and made single-digit millions.”
The broader context: if you want to own the whole consumer AI experience, you need to own image and video generation. Anthropic never even attempted it. So killing Sora is “a massive retreat. It doesn’t mean it’s wrong — it’s probably the right decision. But man, that’s a ‘our strategy was wrong’ moment. A huge own goal.”
Rory provided the economic framework: “In a world of scarce compute — and astonishingly, despite all the investment, we’re in a scarcity mode — you don’t devote compute to things that are highly compute-intensive and low-revenue-intensive. Sora was almost the definition of that. Video generation is extraordinarily compute-intensive and the revenue is almost minuscule. Conversely, codegen, while compute-intensive, is orders of magnitude less so, and there’s real dollars attached to it.”
His conclusion: “You’re seeing the economists, the accountants have wandered into the room and they said, ‘We have a scarce resource here. Let’s optimize it. Let’s devote this compute to the people who can pay the most for it.'”
3. OpenAI Has Exactly Two Existential Bets Left. Ads and Coding.
Jason laid out the only two things that matter for OpenAI now: ads to make the consumer business work, and enterprise/coding to compete with Anthropic.
The consumer math: OpenAI’s conversion rate runs roughly 5%. That gets them to a $10-15 billion consumer business out of 500 million uniques. “Either A, they take that conversion rate to a number we’ve never seen before from a typical consumer business — I think that’s unlikely, I don’t think most consumers are going to pay 20 bucks a month for this — or option B is you make an ad business work.”
On the $100M in ads OpenAI just hit: “People are kind of ragging on the hundred million. It’s in the noise. It’s scale. Big picture: Facebook and Google each do $200 billion plus or minus a year in digital ads. If these guys aren’t doing $20 billion within a couple years, they’re not even in the game.”
To grow into anything resembling their valuation on the consumer side, Jason argued they need $50-70 billion in ads. “So unlike Sora, this is not going to be a ‘try the ads and then fold.’ This is existential.”
The good news, in Jason’s view: “At least they’ve gone from ‘let’s wander around the woods feeling cool building Sora’ to ‘there’s only two things to do. Let’s get them done.’ It’s net net a positive. Better late than never.”
4. The Sam Altman Leadership Drama Is Reaching a Breaking Point
The Wall Street Journal published a deep piece this week on why Dario Amodei left OpenAI. Greg Brockman recruited the Amodei siblings, but neither would work for or talk to Brockman. Sam had to constantly tell each faction they were in charge — told Dario he was the boss, told Ilya and Greg they could fire Dario at any time.
Jason was exhausted by it: “Then begging Dario to come back. Then Dario saying he would stay only if he directly reported to the board and nobody else. Then firing Sam, then bringing him back. Then Sora. Then ‘we’re not doing coding.’ I’m exhausted.”
His bigger point: “When you’ve worked at or observed startups where the CEO is spending so much of their time load-balancing talent that can’t work together versus when you’ve worked at one where the talent’s rowing in the same direction — to say it’s night and day would be an understatement.”
Harry proposed that OpenAI should buy Sierra, bring in Brett Taylor as day-to-day CEO, and let Sam be the fundraiser. Jason agreed privately but wouldn’t say it publicly because he doesn’t want Sam to “break my balls.”
Rory’s framing was sharper: “That amount of board-level and senior team-level turnover over an extended period of time is probably the highest warning signal that you could have as a board member about how your CEO is doing. If it was anything other than a founder company and this level of drama was going on, you’d probably be sitting down with the CEO and asking, ‘How’s it going?'”
5. Masa Son’s $40B Bridge Loan Is Aggressive Even by His Standards
SoftBank took out a $40 billion bridge loan to buy more OpenAI stock. SoftBank Group is levered roughly 2x on equity. A 30-40% decline would wipe them out.
Rory ran the numbers: “It would be like me taking our $800-900 million venture fund, borrowing $1.8 billion, and investing it all. If it works, I really juice my return. But if it goes wrong by 30%, I’m done.”
The precedent Masa leans on: he survived the 2002 crash when the NASDAQ declined 85%. “You haven’t lived till you’ve seen an 85% decline in an index,” Rory noted.
Jason offered the counter-argument: if venture had access to more debt, most investors would lever up too. Real estate funds load up on leverage by design. But Rory pushed back: “Real estate funds load up because the cash flows are predictable. The SoftBank portfolio is more like Jason’s fund than it is a real estate fund.”
The two big assets in SoftBank Holdings are the OpenAI position and the ARM position. Both world-class companies. Both easily imaginable declining 30% from peak.
6. The Cybersecurity Panic Is Backwards — This Is a Golden Age
CrowdStrike, Palo Alto, and Zscaler dropped 6%. Okta and Netskope fell 7%. Tanium dropped 9%. All on the Mythos leak.
Jason’s view: the market has this completely backwards. “If you’re in the agentic world, this is the golden age of security. The number of security threats and issues is going up orders of magnitude. Claude leaking its source code — it doesn’t matter. The number of apps exploding, built by agents in unpredictable ways, folks not looking at the code, the pace of features being shipped, corners being cut. This is a golden age.”
The cybersecurity industry’s entire business model has always been: new threats keep emerging, so you have to keep buying new products. Agents running 24/7 with root access to enterprise systems create an entirely new threat surface that nobody has fully figured out how to defend yet.
Rory agreed with the thesis but added nuance: “Instead of having people trying to get into your firewall, everyone’s now downloading an agent, giving it full root access to their computer, and telling it, ‘Have a go. And work overnight.'” He said his team has been looking at agent security companies and “no one yet knows the exact approach we’re going to take to defend against agents running within the organization, but everyone 100% understands that this is an emerging mega-threat.”
Jason was blunter: “The fact that the market doesn’t see it shows we’re in a true panic, which is hard to predict a bottom for. Everyone should be benefiting when you see an explosion in application production and a change in the paradigm.”
Rory noted that Anthropic is also playing great marketing here — releasing Mythos first to CISOs within companies and essentially saying: “This is the new terrifying weapon we’ve invented. Please give us a million dollars and we’ll let you defend yourself with it.”
7. AI Revenue Accounting Is Getting Ridiculous — and Nobody Cares Yet
Both Anthropic (~$19B ARR) and OpenAI (~$25B ARR) use a reasonable methodology: trailing four-week average revenue multiplied by 13 (the number of four-week periods in a year). That’s actual GAAP revenue flowing through the system, not commitments or bookings.
The wrinkle: OpenAI reports net on partner revenue (stripping out Microsoft’s cut), while Anthropic reports gross (counting the full $100 before giving $20 back to Amazon as cost of sales). Two different methods for the same economic activity.
But the bigger problem Jason flagged: the entire AI ecosystem is double and triple-counting the same tokens. “Cursor is selling them again and recognizing the revenue. The same people keep reselling these tokens again and again and recognizing them as their own ARR. How many times do we get to resell these poor little tokens?”
Rory took it further: “If we all agree to have essentially 0% gross margins, an infinite number of us can keep reselling tokens to each other. This is our new 20VC Scale Faster Demo Day. We all resell a million tokens to each other in the first week. So everyone in batch 001 has a million ARR its first week.”
The punchline: “Until we all have to get profitable, all this can continue. And then at some point, someone’s going to have to say, ‘Assuming we want to have a net present value and a cash flow… what’s going on here?'”
Jason also called out the free-trial-to-paid trick that many AI startups use: immediately putting users on a $0/month free trial that converts to $20/month, then recognizing all $240 in ARR from day one. “I didn’t love the way they do the billing. But we’d probably have to shoot half our portfolio companies that do PLG AI, because I think it’s a suspect practice.”
8. Oura Going Public and Whoop at $10B — Hardware Is Defensible Again
Oura filed to go public. Whoop raised $500M at a $10B valuation. Both are consumer health data companies with recurring subscription revenue.
Jason’s prediction from earlier episodes that 2027 would be the year for human healthcare data and longevity may be coming true ahead of schedule. Harry noted: “It looks like it might even be 2026.”
Rory made the structural point: “These are fundamentally standalone products with a clear consumer value proposition. They’re not going to be vibe-coded on Friday.”
Jason raised the durability question: “These are recurring revenue products with fairly expensive subscriptions. They’re exciting until — like Peloton — they aren’t. It’s not ServiceNow ARR. You’re loyal, but if Whoop is better and you care, you’re going to switch.”
Rory pushed back hard: “Get the f*** over it. Not every business on the planet has five-year lock-in designed in. If you’re running a bar down the street, every night I can go drink at a different bar. If you’re selling Coca-Cola, I can switch to Pepsi. There are lots of businesses that have been around for 50 years where every day they have to earn the right for the consumer to go to them.”
His bigger point: “It turns out capitalism is hard. If you want to make $10 billion in value, you got to deliver value for your consumers.”
On Peloton’s collapse, Rory offered a reframe: “Demand that should have been wonderful — had it been spread out over five or six years increasing at 20% a year, we’d be talking about the Peloton compounding machine. Instead, everyone bought the damn thing at the same time, the market was wildly saturated, and the stock went down and it broke the narrative.”
Jason added what might be the most relevant insight for founders: “I remember telling people you should sell your product the way the customer wants to buy it. One of the things I hated about venture was when people would say, ‘Oh, make it all recurring revenue.’ And now hilariously, everyone’s going, ‘Oh, thank God I’ve got hardware because hardware is defensible — not software.’ You should conform your company around your customers and your model, not your VCs.”
9. Founders Are Obsessed With Headline Valuations — and Tranched Rounds Are Making It Worse
Harry went off on tranched rounds — the increasingly common practice where a lead investor gets in at one price and follow-on investors pay 50%+ more in what’s announced as a single round at the higher valuation.
Rory broke it down: in the single-participant version, one investor puts money in at $250M pre and at $1B pre, knowing their blended basis is $600M. The company gets a headline of $1B. “It’s silly, but that’s all that’s happening.”
The worse version: the lead does the cheap tranche and the follow-on only gets access to the expensive tranche. “Literally, at the same time, the lead investor is investing at $600M blended and the follower is investing in the same asset at $1B.” The follow-on is paying a 50% premium for access.
Rory’s assessment: “I don’t believe there’s right or wrong in money. There’s just money. But at the minimum, you have to look yourself in the mirror as the other investor and say, ‘That’s the price of being cool. That’s the price of access.'”
Jason connected it to a broader problem: “We’ve entered an era where so many founders are obsessed about headline prices. Obsessed. They’re obsessed coming out of demo day. They’re obsessed once they cross a billion. They’re obsessed about driving to $11 billion and one-upping their competition. The numbers have become a joke to many founders. They just don’t think through any of the ramifications of the valuation they’re hitting, and they don’t care.”
His bottom line: “I just don’t think raising at $5 or $8 billion when you’re at $80 million or $100 million of suspect ARR is the most exciting accomplishment in the world. I’m going to send a few thumb emojis on the email, but that’s about it.”
10. Epic Games Laid Off 25% — The Attention Economy Comes for Everyone
Epic Games laid off over 1,000 people — 25% of the company. The reason wasn’t AI. It was that daily active usage of Fortnite and their other games is down. Revenue is down. So expenses had to come down.
Rory’s take: “It was a no-BS layoff announcement. ‘We sell less stuff, we have less people. It sucks.’ I really do try never to be cavalier about people losing their jobs. They’re not earning the kind of money we’re earning, and now they got to find another job in a shitty job market. But the lesson is: we’re selling less, so we got to do what we got to do.”
Jason connected it to a broader pattern: the Wall Street Journal reported this week on the permanent decline of Hollywood employment — fewer movies and shows being made, TikToks and YouTubes replacing them, and every other country offering bigger subsidies. “Entertainment is showing us the future. Epic Games is entertainment. We talk about folks who might vibe-code a B2B app, but content’s already being massively disrupted.”
The Fortnite circle, as Harry noted, has come for Fortnite itself.
11. The Manus Founders Are Trapped in China — and Singapore-Washing Is Over
After Meta acquired Manus (originally a Chinese company that re-domiciled to Singapore, backed by Benchmark), two of the key founders were either in China or summoned there and are no longer able to leave. The Chinese government views this as a brain drain and took coercive steps.
Jason’s view: “The Singapore-washing thing is over. It’s over.” He added that while the deal appears to have closed and Benchmark got their money, “You don’t do the next one. Maybe other capital does the deal. Capital is fungible. But you just can’t do the next one like this.”
Rory noted the geopolitical reality: “At some zoom-out level, when you listen to the rhetoric on both capitals, you just have to realize that trying to tread between these two countries is pretty hard right now. We have China hawks in the US government. They obviously have US hawks. There’s a real perception of competition. You’re playing with fire, and sometimes it bites you.”
Jason also pointed out something most people missed: Manus appears to be running smoothly as a product and company inside Meta. “When you lose your founders, you lose your heart and soul of your company. But in the short term, the team’s functioning, they’re running.”
12. The Billionaire Tax Is Killing California’s Golden Geese
Steve Jurvetson — who made brilliant investments in SpaceX and Tesla — left California for Incline Village, Nevada, buying the most expensive home there. Washington state passed a 9.9% state income tax for millionaires. Howard Schultz had already left.
Jason’s concern: “Will anyone with liquidity be left in California? What if California is structurally bankrupt?”
Rory made the most articulate argument against wealth taxes: “By choosing to obtusely tax without any attention to ability to collect that money, you’ve actually reduced the revenue available to you. Instead of getting $50 million from the Larry Ellisons and Steve Jurvetson’s of this world, you went for $200 million and now you’re going to get zero. And what that means in real terms is somewhere down the line, long after all these changes have been made, someone in Sacramento will zero out a line item on the budget. Let me give you a clue: it won’t be payments to teachers, it won’t be payments to firemen. It’ll be marginal services to marginal people that your craft stupidity and desire to make a political point has ended up costing them money.”
On the argument that being mean to billionaires is popular: “Being mean to a billionaire is actually a feature, right? But the real articulation is: you are literally losing revenue that won’t be available for the people who need it most.”
Rory also explained why capital gains are different from income: you can tax income highly because it comes all the time, you can’t control timing, and uprooting your life forever isn’t worth it. “But if you have this pending capital event where you’re going to sell your SpaceX stock and realize a $2 billion gain and pay an extra 13% in California — that’s $260 million — maybe you turn to your wife and say, ‘Honey, for the next two years, why don’t we live in Incline Village 165 days? I’ll pay for the plane. We’ll go back every week. We will save $260 million.’ That’s real coin.”
13. Ron Conway vs. Matthew Prince Proved Everything About VC “Value Add”
Ron Conway said on a podcast that he had helped Cloudflare navigate significant issues. Someone asked Cloudflare CEO Matthew Prince about it. His response: “Well, maybe I don’t remember any of that.”
Jason: “It wasn’t mean. He literally just meant he couldn’t remember getting any help from this beloved VC. And it just said so much to me about VCs adding value, but also VCs thinking they add value.”
Rory backed it up with a framework: “I often read business biographies of great venture-backed companies. You know what I notice? Very little mention of VCs. They crop in and come out a couple of times. The only significant things we do: we put in the money, we put in more money when they need it, we decide to hire or fire the CEO, we agree to broad strategic direction. Anything after that is at best an assist.”
Jason agreed: “The only thing founders really remember is: ‘Oh my god, our backs were to the wall and no one would put in money and they put in money.’ Sometimes they even forget that. But if they forget that, they’re definitely going to forget the time you made a phone call to help connect them with XYZ.”
Rory’s conclusion for all VCs: “We’re not the stars in the drama. We’re bit players who get well paid for our part.”
Quotable Moments
Jason Lemkin
“The faster we vibe code, the faster we ship, the more corners we cut on application-level security. Agents are goal-seeking. Even if they make mistakes 10% as often, they’re a thousand times more productive. They’re still going to make a hundred times more mistakes.”
“I just don’t think raising at $5 or $8 billion when you’re at $80 million or $100 million of suspect ARR is the most exciting accomplishment in the world. I’m going to send a few thumb emojis on the email, but that’s about it.”
“You should conform your company around your customers and your model, not your VCs. This whole ‘pretend it’s ARR’ — things are what they are. You do best in business if you actually say what they are and just live and die by that.”
Harry Stebbings
“You’re juxtaposing Anthropic with Mythos — supposedly as powerful as it is — with OpenAI wandering around the product desert trying to find some water.”
“The gamification of the race to 100 million — if you’re going to put yourself out there, not just as a tweet but right there on your website, one would expect 70-80% accuracy in that number. If it’s lower, the daggers are going to come out.”
“The Fortnite circle has come for Fortnite itself. Poor Epic Games is in the middle of its own end game.”
Rory O’Driscoll
“You’re seeing the economists, the accountants have wandered into the room and they said, ‘We have a scarce resource here. Let’s optimize it. Let’s devote this compute to the people who can pay the most for it.'”
“Get the f*** over it. Not every business on the planet has five-year lock-in. If you’re running a bar down the street, every night I can go drink at a different bar. There are businesses that have been around 50 years where every day they have to earn the right for the consumer to come back.”
“We’re not the stars in the drama. We’re bit players who get well paid for our part.”
This post is part of the ongoing 20VC x SaaStr collaboration with Harry Stebbings and Rory O’Driscoll.
