# 20VC x SaaStr: Pacing the Frontier, Meta Ships Muse And It’s Great, and Miro Sells for $1.35B After a $17.5B Mark > Source: > Published: 2026-09-20 14:10:43+00:00 # 20VC x SaaStr: Pacing the Frontier, Meta Ships Muse, and Miro Sells for $1.35B After a $17.5B Mark *Plus: a $1B round at $10B for Instinct, Discovery Loop from $10B to $50B in weeks, and Mistral’s €3B sovereignty round* ## #1. Dario Called for Pacing the Frontier, and the Market Moved 0.1% Dario Amodei published a call to pace the frontier, proposing an external body that would monitor and constrain frontier model capabilities. Sam Altman agreed. Elon agreed. The letter names three risks: cyber, economic disruption, and losing control of the models. Separately, an Anthropic safety lead put a 10% number on catastrophic outcomes, and that’s the number everyone repeated. The reception was close to universally hostile. **Rory’s read:** two of the three risks don’t hold up. The cyber risk is real, but that capability is already out and in the open-weight ecosystem. The economic argument fails on its own terms, because by that logic we’d still have 73% of the workforce on farms. Only loss of control is genuinely hard, and that one has no good answer. On implementation, the path runs from unlikely to impossible: voluntary third-party monitors, then mandatory ones, then every democratic government agreeing, then China agreeing. **Jason’s read:** this is an S1 risk factor being drafted live. Anthropic is going public. Get the debate out of the way now and at a $2 trillion IPO it’s a non-issue on the roadshow. That’s the CEO’s job. **David Sacks and Lina Khan agreed with each other,** which doesn’t happen. If the risk is that material, fixing it is the CEO’s job, and shutting it down is on the menu. Jason’s add: expect congressional hearings on this for the next 24 months. Relatives who have never texted about tech are texting about tech. **Market reaction:** a one-day markdown on semis, a roughly 10% jump in cyber names like CrowdStrike, and software outperforming semis on the day. Capitalism priced a 10% extinction claim at about a tenth of a percent. ## #2. Meta Shipped Muse, and It Costs $3 to $4 Per User to Deliver Meta launched Muse, its consumer AI assistant, running on Meta’s own model. Every user gets a free VM: roughly two CPUs, two GPUs, 8GB of RAM, 100GB of storage. Jason had it send the show’s story list on schedule and rebuild saastr.com through WordPress. It worked. **Jason’s read:** cost is the advantage. Replit, Lovable, Vercel and the rest pay something like $3 to $4 per user to deliver that kind of environment, and they work that number down every week because it’s a large part of COGS. Wix, before Base44, served a site for about two cents. Meta already owns the infrastructure and the model, so it hands out more compute than anyone and runs faster doing it. **On how it got built:** this went from zero to launch because Zuck moved a large chunk of the AI org onto it the moment OpenClaw took off. Not 20 engineers, closer to 500, responding to bug reports on Twitter on a Saturday night. ## #3. Does an AI Assistant Need a Killer App? **Jason’s position:** no VisiCalc yet. The best use case Meta itself demos is scheduling a daughter’s carpool. For this category to matter it needs to be open eight hours a day the way Claude Code or Codex are, and nobody has shown that yet. OpenClaw never had a killer app either. **Harry’s position:** it may not need one. Instinct handles his bookings, travel, restaurants, shopping and calendar invites. It’s incrementally better at a long list of small things, and that might be enough. The real complaint is latency: minutes per response, which feels like the first weeks of ChatGPT. **Rory’s read:** Facebook already tried the super app with Messenger and it failed, because booking a flight in a chat window was worse than just loading the United site. Intelligence in the backend changes that math. So yes, it’s a category. Whether a standalone can win it is a separate question. ## #4. Jason’s Investment Committee: A $1B Round at $10B for Instinct Instinct is reportedly raising $1B at a $10B valuation. The round history: roughly $50M pre in April, $500M pre with Kleiner shortly after, then $2.5B, now $10B. **The case for, as Jason argued it to the partnership:** Meta can’t go cross-platform. Muse is anchored to Meta’s own surfaces, it won’t work across carriers and services, and Facebook itself is your grandmother’s app. Meta’s attention also fades on products that generate no revenue. Workplace had the highest NPS of anything at Meta and still got abandoned. And the Instinct team is as strong as anything he’s seen. **Jason’s actual vote: no.** For the entire AI cycle, app companies got a hall pass because the labs didn’t build apps. Meta is now building the app, with its own model, its own compute, and structural cost advantages. Subsidizing $5 to $10 per user per month with venture dollars works until 10 million users turns into a $1.2 billion annual nut and the next round doesn’t clear. **Rory’s read:** the risk/return was wildly attractive at $50 pre, good at $500, thinner at $2.5B, and at $10B you need a $50B outcome to make it work. The comp is Cursor, now closed into SpaceX in an all-stock deal at an implied $60B, the largest software M&A outcome on record. Cursor was at roughly $2B in ARR and growing faster than any business software company ever measured when that price got set. You don’t get a $40B exit without enormous revenue underneath it. Poolside showed the talent-and-assets path exists, and with Brett Taylor in the founder’s orbit there’s an obvious buyer. But underwriting a deal where an M&A outcome is the only plausible exit is a bad habit, because acquisitions are capricious. **Harry’s read:** every reference call comes back with the same description of Noah Shin as a generational talent, and generational talent in a strategically attractive category gets bought by multi-trillion-dollar acquirers. A $50B to $60B upside case isn’t crazy. ## #5. Menlo Is Underwriting to 100 Exits North of $25B Jason walked through the framework behind Alex Kurland’s move to Menlo Ventures: the firm is modeling roughly 100 future outcomes worth $25B or more, and building the portfolio to catch some of them. There are about 81 companies above $25B today, up from 23 ten years ago. The live example landed the same week. Jeff Dean’s Discovery Loop, founded with Sanjay Ghemawat, Quoc Le and Oriol Vinyals after Dean left Google in August, raised $1B at roughly $10B, and is now in talks at about $50B. No disclosed product. Five times the mark in a matter of weeks. **Jason’s read:** this is why $10B entry prices pencil for some firms and not others. If the good outcomes are all north of $25B, the new decacorn is $25B, and a $10B round is a 3x. If you don’t believe there are 100 of them, the same round is indefensible. **Rory’s read:** venture, unlike PE, was never about valuation. You’re either at the head of the train where everything is possible, or at the tail where nothing is. ## #6. Miro Sold for $1.35B After a $17.5B Mark Bending Spoons acquired Miro for $1.35B. The last round was 2021 at $17.5B. Miro is doing roughly $600M in ARR, growing high single digits, and is cash flow positive, so the deal cleared at about 2.7x revenue. Accel made money. Founders and employees made money. The late-stage investors got roughly 1x. Part of the consideration rolled into Bending Spoons stock. **Rory’s read:** this was the largest stale valuation left from the 2021 cohort, and the cleanup was inevitable. Late-stage math still works here: if your losers return 1x on preference, the distribution is net positive by definition. A 1x is a bad deal, not a wipeout. **Jason’s read:** when a company sells for 2.4x or 2.5x, you know there was no second bid. Anyone could have paid 2.5x. Salesforce, Thoma Bravo, Francisco Partners. Nobody did. ## #7. Bending Spoons Screens 1,000 Targets and Buys 5 to 10 Bending Spoons looks hard at about a thousand targets a year and closes five to ten. Their CEO’s public position is that he doesn’t care what you founded ten years ago, he cares what you’re doing now. **Rory’s read:** selling at 2.7x to a buyer trading at 12x to 14x is an admission. It says a five-party venture syndicate can’t do what one owner will do: cut extraneous cost, raise prices, accept a fair amount of churn, and keep going. That isn’t venture DNA. Some assets are better owned by a single owner. **Jason’s read:** study the usage graphs on their acquisitions and the churn is brutal. They are not revitalizing these products. They are raising prices and cutting cost. If you’re a customer of one of these companies, get ready for the 40% increase. **Rory’s follow-on:** what Bending Spoons is measuring is marginal propensity to pay. The industry overinvested in sales and marketing and sold a lot of customers who had to be sold. Their model says: I don’t want the customers who had to be sold. I want the ones who resent the price increase and stay anyway. ## #8. Two Chairs Left in Duck Duck Goose Airtable got one offer. Miro got one offer. PE is mostly sitting out, Thoma Bravo is mostly sitting out, and even Bending Spoons doesn’t have an unlimited balance sheet. **Jason’s read:** this is the era of capitulation. Miro and Airtable are decent assets, roughly $600M in ARR, still growing, cash flow positive. If that’s the best outcome those two can get, anything weaker doesn’t get a chair at all. Those companies go into zero to 5% growth mode and nobody buys them. If you have a seat available, take it now. ## #9. Matt Mullenweg Was Out at Automattic for a Day, Then Back Automattic’s board moved to replace Matt Mullenweg. He was out, then back as CEO, and the independent directors left. **Rory’s read on the mechanics:** the board almost certainly had the votes to replace the CEO, and the founder had the votes to replace the board, so he did. This wasn’t a board full of evil VCs. It included real independents who didn’t need the grief. Meanwhile the fight is over a shrinking prize, because you can now build in Replit or Lovable what you used to build in WordPress. **Jason’s read:** Automattic would have been a great company if it had never raised venture capital. A bigger Basecamp, $500M in revenue throwing off $200M, run by 80 people, with the founder doing exactly what he cares about. And venture-backed open source requires ruthlessness. Matt left hosting to WP Engine because it looked like a commodity, and left commerce short of what Shopify built. Now he wants that revenue back and it isn’t his. ## #10. Mistral Raised €3B, Led by Samsung Mistral raised €3 billion, the largest tech round in European history, led by Samsung. The prior round was led by ASML. The company is tracking toward roughly $1B in revenue by the end of the year. **Rory’s read:** this is a sovereignty purchase, not an entry in the frontier race. Europe watched US export controls cut off access to frontier models with no carve-out for the UK, France or Germany, and concluded it can’t depend on American labs. The precedent is Airbus, which took decades and eventually passed Boeing. If US winners are worth a trillion and European GDP is 70% to 80% of American GDP, a $30B to $40B European winner is a reasonable end state. **Jason’s read:** when Samsung leads the round and ASML led the last one, the valuation deserves an asterisk. The money is real. The price isn’t a market-clearing price set by financial buyers. ## #11. Adobe Named a New CEO After a Year of Deliberating Adobe promoted an internal candidate after a year-long succession process. Roughly $25B in revenue, roughly $105B in market cap, trading at a sub-10x cash flow multiple. AI usage metrics went up. Net new ARR went down. **Jason’s read:** nothing changes here. This is capitulation with enough scale to survive it, which is what Miro didn’t have. Adobe grinds at six to nine percent with heavy free cash flow, and in three years the market cap is roughly where it is now. **Rory’s read:** above 30% growth you get valued on revenue and everything is forgiven. Below 30% you get valued on EBITDA and nothing is. Canva is growing 20%, down from 30%, per Australian filings. That’s still good, and it’s still deceleration. Box took three or four years to get through that transition as a public company, because going from 6x or 7x revenue to 20x cash flow means growing cash flow 30% a year just to hold the stock flat. **Jason’s add:** doing that transition while private is worse, because you have to get through the window to get liquid. Stripe avoided it by reaccelerating. Most don’t get that option. ## Instinct at $10B and Miro at 2.7x, in the Same Week The two ends of the market priced themselves within days of each other. A four-month-old assistant company went from $50M pre to a reported $10B, Discovery Loop went from $10B to $50B in a few weeks with no product, and a $600M-ARR collaboration leader with positive cash flow sold for 2.7x to the only bidder who showed up. Nothing in between got priced at all. If you’re running a B2B company today, that gap is the strategy question: you’re either underwriting to the head of the train, or you’re managing an asset that needs to be cash-generative before anyone asks you what your growth rate is. ## 3 Quotable Moments From Each **Rory O’Driscoll** 1. “If the feds really thought that there was someone in downtown San Francisco building a technology that had a 10% chance of blowing up the world in the next 10 years, they would move in with a SWAT team, kill everyone in the place, and close it down.” 2. “If you’re building something that you can’t control, then maybe you should stop building it. You’re the CEO.” 3. “The great thing about the late-stage business is this: if your losers give you a 1x, then you’ll die rich.” **Jason Lemkin** 1. “This was just a risk factor in an S1 done live. Anthropic’s going public, and he’s getting ahead of it so that when the $2 trillion IPO happens, it’s a non-issue.” 2. “This is the threat every VC worried about. We all got a hall pass since the start of AI because the LLMs didn’t build any apps. This is the one they’re building.” 3. “I’ve ended the game of duck duck goose for me. I’m not running around the chairs anymore.” **Harry Stebbings** 1. “I don’t know if it needs a killer app. It does all of my bookings, travel, restaurants, shopping, calendar invites. It’s just incrementally better than everything else.” 2. “I did an Instagram reel on it. I had over a thousand DMs asking for invite codes. I’ve never had a thousand DMs on the back of a reel.” 3. “I was with one of the biggest CIOs the other day and asked if he’d ever seen a time like this in 30 years. He said he’d never seen a time like this. This is more nuts than it’s ever been.”