20VC with Harry StebbingsInstinct Raising $1B at $10B & Meta Launches Muse | Miro Sells for $1.36B After a $17.5B Valuation
20VC with Harry Stebbings
Pacing the Frontier Debate Begins 0:00
The discussion opens with Dario Amodei's call to "pace the frontier," a statement about slowing down AI capability development that Sam Altman and Elon Musk both agreed with, including the idea of some external regulatory body to oversee model providers. Rory pushes back hard, calling the reception rightly hostile. He argues that if the federal government truly believed there was a 10% chance a technology in downtown San Francisco could cause human extinction within ten years, they would already have shut it down with force, the way they would treat someone building an unsafe nuclear reactor. He separates Dario's own stated concerns, cyberattacks, economic disruption, and loss of control over AI, from the more extreme "10% chance of extinction" comments made by an Anthropic employee who had quit, arguing these are very different claims that got conflated.
Jason Reframes It as IPO Strategy 6:01
Jason offers a more tactical read, suggesting Dario's comments function as a risk factor being aired ahead of Anthropic's anticipated IPO, so that when the company goes public at a potential two trillion dollar valuation, the concern is already priced in and debated publicly. He predicts this will trigger two years of congressional hearings and public anxiety, noting he personally received texts from relatives outside the tech world asking if AI was going to kill everyone. He connects this to why he believes Trump quickly dismissed the issue rather than let it dominate the conversation.
Sacks and Khan Both Push Back 9:00
Rory highlights David Sacks' comment that if the risk is real, it is Dario's job to fix it or shut the company down, comparing it to product liability, since a company cannot be allowed to kill people with its product. He notes the unusual alignment between Sacks and antitrust regulator Lina Khan on this point. The group agrees that of the three risks Dario raised, job losses and cyberattacks are either overstated or already unavoidable, but the idea of losing control over recursive self-improving AI is the one genuinely hard problem, and the point stands that if you cannot control what you are building, you should stop building it rather than ask the world to regulate you instead.
Every Technology Has a Dark Version 16:00
A tweet is quoted making the point that every positive AI capability has a corresponding dark version, superhuman coding ability means superhuman hacking ability, drug design skill means poison design skill, and structural engineering skill means the ability to find flaws that bring buildings down. Without real guardrails, especially in open weight models, the dark version can be just as capable as the light one. The comparison is drawn to how the Catholic Church resisted the printing press and how authoritarian regimes resist the internet, each new technology carrying both benefit and harm that society learns to manage over time.
Markets Barely React 17:31
Despite the extinction talk, markets moved only slightly, semiconductor stocks dipped mildly while cybersecurity stocks like CrowdStrike jumped about 10 percent, reflecting where investors saw real near-term risk. The panel jokes that capitalism absorbed a stated 10% extinction risk with a 0.1% market move. They note that funding rounds have become so large that two billion dollars is now considered a modest starter round, and observe that neither Sam, Dario, nor Elon own large equity stakes in their companies, suggesting they aren't pure profit maximizers, though the bigger worry is the estimated 10 to 15% of founders generally who are described as sociopaths capable of misusing the technology.
Meta Ships Muse After OpenClaw 21:00
Attention shifts to Meta's release of Muse, an AI assistant product that Jason reveals was built by around 500 engineers working intensely since OpenClaw launched, not the smaller team the hosts had joked about earlier. Jason describes Muse as excellent software that handles tasks like sending emails, making reservations, and rebuilding a WordPress site smoothly. He explains that Meta has a major cost advantage because it already owns the infrastructure and its own LLM, called Muse LLM, so it can offer generous free compute resources that cost competitors like Replit or Vercel three to four dollars per user to provide. Despite the technical polish, the discussion ends on an open question about whether Muse has a genuinely compelling use case beyond routine tasks like booking a table at a restaurant.
Debating Whether Muse Needs a Killer App 25:02
The hosts question whether Meta's Muse or the startup Instinct actually need a single defining "killer app" to succeed, the way spreadsheets once launched personal computing. One host argues that Mark Zuckerberg's own examples for Muse, like scheduling his daughter's carpool or ballet lessons, sound trivial and unconvincing for something built by a trillion-dollar company. Another counters that Instinct already handles his travel bookings, restaurant reservations, shopping, and calendar invites reasonably well, even if latency is a real problem, with responses sometimes taking minutes to arrive. The group settles on three standard venture questions to test the category: is this really a category, who wins it, and are investors being paid enough for the risk.
The Case For Backing Instinct 29:01
The conversation turns to a hypothetical billion-dollar round for Instinct at a ten-billion valuation, framed as an investment committee exercise. The bull case rests on Meta's structural limits: Muse won't work across other platforms or carriers, and Meta has a history of losing focus on non-core products, citing Workplace, which had strong internal satisfaction scores but was eventually abandoned once its champion left the company. The Instinct team is described as unusually talented, with references to founders who are ranked among the top players in League of Legends and World of Warcraft, and the recommendation is to lead the round while staying cautious with reserves for future rounds.
The Case Against Backing Instinct 31:32
The opposing view is that Meta owns the large language model, the compute, the cost advantages, and the speed advantages, and is now actually building a full consumer app rather than stopping at half-measures like Claude's design tool. This is described as the exact threat venture capitalists have worried about since the start of the AI boom. Instinct's slowness is read as a sign of high compute costs, which is why it needs to raise a billion dollars, and there's a risk it could end up like Poolside, a company with millions of users but a cost structure it can't sustain without repeatedly raising more venture capital.
Meta's Incentive And The Poolside Parallel 34:31
One host argues Meta has every reason to go all in on this, since spending over a hundred billion dollars on AI without prioritizing a personal assistant product would be strange. Even so, Instinct could still succeed the way Poolside did, not through profitability but through acquisition, since Poolside ran out of capital yet delivered a strong outcome when Nvidia bought its model and talent. Instinct's founder, Noah Shin, is described by people who know him as one of the most talented founders around, comparable to figures tied to Sierra and OpenAI chairman Bret Taylor, which raises the possibility of an eventual acquisition in the fifty to sixty billion dollar range.
Weighing Venture Bets On Acquisition Outcomes 36:30
One host says he was raised to avoid bets that require an acquisition to succeed, since such outcomes are unpredictable. Still, he acknowledges the expected value across a portfolio of such bets could be positive even with high variance, and notes Kleiner's earlier round in this company was done at five hundred million, a stage he considers more reasonable. Tracking the deal's progression from fifty million in April to five hundred million, then two and a half billion, and now ten billion, the group agrees the risk-return profile was far more attractive earlier and has grown steadily less appealing.
Miro Sold To Bending Spoons 41:31
The discussion shifts to Miro, the interactive whiteboard company once valued at 17.5 billion in 2021, which sold to the Italian acquirer Bending Spoons for 1.3 billion. Analysis shows early investors and employees still made money, but later-stage investors got roughly a one-times return. The deal is called inevitable, since Miro's valuation had been stuck since 2021 while headcount kept growing and comparable companies raised at ten to twenty billion. Bending Spoons' CEO reportedly said the firm doesn't care what a founder built years ago, only what they're doing now, reflecting its approach of buying stalled SaaS companies, cutting costs, raising prices, and accepting churn to reach profitability, a strategy venture-backed boards are often too fragmented to execute themselves.
Few Seats Left For Struggling Unicorns 48:02
The hosts compare the remaining exit opportunities to a game of duck duck goose with almost no chairs left. Bending Spoons reportedly reviews about a thousand potential targets a year but only completes five to ten deals, and most private equity firms are largely sitting out these SaaS acquisitions. Miro and companies like Airtable, which also reportedly received only one acquisition offer, are seen as relatively strong assets still growing and cash flow positive, meaning weaker companies likely won't find buyers at all and will instead drift into low or no growth with no one interested in acquiring them.
SBF's Supreme Court appeal odds 51:00
The hosts discuss Sam Bankman-Fried's chances of getting his case heard by the Supreme Court. One host argues the appeal centers on constitutionality, not sentencing guidelines, noting that under bankruptcy court terms, creditors were technically repaid in whole with interest, even if they could have made more money otherwise. He predicts the Court will take the case, though getting a hearing is far from being freed. He also notes SBF apparently did not enrich himself materially, which he calls an interesting wrinkle in the case.
Debate over punishment for financial crimes 53:00
One host argues white collar crime should be punished, comparing SBF's 30-year sentence to a Goldman Sachs figure who misallocated brokerage funds years earlier and walked free. He calls it a shitty world where someone stealing $20 goes to prison while someone stealing $10 billion doesn't, because of race and class. He also argues that taking money without permission is wrong regardless of investment skill, since the person whose money was taken didn't consent to the risk, and punishment deters future behavior.
Matt Mullenweg reclaims WordPress control 55:00
The hosts turn to Matt Mullenweg, founder and CEO of Automattic, who was briefly ousted by the board before reversing the move and returning as CEO. They describe a company whose stewardship of the open-source WordPress project has grown troubled, especially amid Mullenweg's fight with WP Engine, a hosting company he's accused of profiting from the ecosystem without giving back. One host suspects Mullenweg used bylaws to effectively vote out the board that tried to remove him, after which independent directors resigned rather than fight a losing battle.
Venture money versus lifestyle business tension 1:00:00
The hosts argue Automattic might have thrived as an unfunded, founder-run business rather than a venture-backed one, comparing it to Basecamp or 37signals, companies happy growing modestly while pocketing tens of millions a year. They suggest raising heavy venture capital saddled Mullenweg with growth expectations that don't fit his mission. One host adds a second dimension beyond venture-versus-lifestyle: whether a company is a good steward of an open-source community, arguing Automattic hasn't been, though as full owners they're entitled to run it however they choose.
Mistral's €3 billion raise and Europe 1:04:00
The hosts note Mistral just raised €3 billion, Europe's largest-ever tech round, led partly by Samsung, with the company projected to hit a billion in revenue by year's end. They argue the deal isn't about competing at the AI frontier but about sovereignty, pointing to reports that the US pressured Anthropic to cut off other countries from a model called Fable, which convinced Europe it needs its own AI champion. They compare it to Airbus, built decades ago so Europe wouldn't depend on American planemakers, eventually overtaking Boeing.
Adobe's leadership change draws skepticism 1:09:30
The hosts discuss Adobe naming a new CEO after Shantanu Narayen's yearlong succession process, calling it a nothing-burger signaling no real strategic change. They note Adobe's net-new annual recurring revenue actually declined even as its AI-linked revenue metric rose, meaning growth is being reshuffled rather than created. With a market cap near 105 billion and trading near ten times cash flow, they expect Adobe to stay roughly flat for years, much as it did before Creative Cloud unexpectedly reinvigorated growth, unless a similarly unexpected AI breakthrough emerges.
Canva's growth is decelerating 1:15:30
The hosts note that Canva's revenue growth, tracked through Australian financial filings, has slowed from roughly 30% to about 20%. They frame this as a real deceleration rather than a stall, and debate whether Canva now counts as a growth stock or a value stock heading toward discounted cash flow analysis.
Growth versus value valuation 1:16:00
One host cites an insightful Twitter post arguing that above 30% growth a company gets valued on a revenue multiple, while below 30% growth it gets judged on an EBITDA multiple instead. He recalls Box going through this exact shift as a public company, taking three or four years to rebuild cash flow enough just to keep its stock price flat.
The painful transition to liquidity 1:17:30
He explains that moving from a growth valuation to a value valuation is brutal whether a company is public or private, and doing it privately is especially hard because it delays reaching a public offering. Stripe managed to avoid this trap by reaccelerating its growth, and he hopes Canva can do the same, since sitting at high cash flow but slowing growth forces investors to choose between liquidity now or a bet on renewed acceleration.
AI-generated summary. It can be wrong or incomplete - check anything that matters against the original.
