𝗠𝗮𝘁𝗵𝗲𝘄 𝗕𝗲𝗻𝗷𝗮𝗺𝗶𝗻 🇦🇺

Do my agents need better judgment, or should they run my analytics?

Another Aussie company passed up by Australian VCs went on to raise $4m seed from @foundersfund. Not massive by today’s standards, but massive if you know the backstory of @paul_conyngham at @gamgeevetmed This is becoming more than a coincidence, it’s becoming a pattern. By my count ~40+ companies have been missed by Australian VCs including @airtreevc @blackbirdvc and @SquarePegCap. The real question to ask is why? Why do so many Australian founders get passed by Australian VCs who then go on to raise from US VCs?

Me running my orchestra of agents from the terminal:

wow. this is super super good. thank you for writing this. one thought: what happens when agents start writing the kernels and new model architectures keep appearing? I think a lot of today’s chips become meh. they were designed by humans for humans. brute-force approaches that use massive memory and compute in ways that only make sense when AI (ie not humans) are doing the mapping will emerge is my bet. memory wall matters but gets solved differently.

The AI Bubble will create Roadkill @DaFrankel (David Frankel), Managing Partner, @fcollective (Founder Collective), interviewed by @HarryStebbings (@20VC) Summary: David Frankel has written seed checks into Uber, Coupang, PillPack, SeatGeek, Whoop, Shield AI and Suno across 18 years at Founder Collective, and he expects this AI wave to leave more wreckage behind it than any wave he has seen. The money at the top of the market is competing for a handful of trillion-dollar names while the returns math sits far below that, in companies nobody is bidding on yet. His case is for getting in early, staying patient, and holding your frameworks loosely enough that they don't cost you the deal of the decade. 1. The $2.6 Billion Median. Fewer than 100 companies created in the last 25 years are sustainably worth over $10 billion. The median of the top 500 companies from that period is $2.6 billion, and owning 5% of one of those returns a seed fund. Frankel says venture has narrowed into a pyramid because Cambridge Associates and the funds of funds are selling access, and a fund that misses the $3 trillion names gets hard to sell. The returns math works at a fraction of that scale, which is why he says you don't have to be in the one. 2. Hollywood Odds. 95% of the companies riding this wave will not be there. Frankel believes OpenAI, Anthropic and SpaceX are the Metas and Googles of this era, and that the wreckage around them will be enormous. He has heard "this time is different" from Harry Stebbings and everyone else for 11 years, across the internet, SaaS and mobile waves that produced fewer than 100 companies over $10 billion between them. Asked whether another crash is coming, his answer is definitely, and nobody knows when. 3. Everyone Is AI Now. Saying you're not an AI company today is like saying you don't use the internet. Frankel's live example: a founder who spent 20 years in SAP consulting, built a platform, knew which part of it was still bad, started playing with Claude Code alongside his CTO, and came back raising at a $20 million cap. Founder Collective does those deals all day long. The AI label stopped carrying information, so what matters is what the founder knows that nobody else does. 4. The Insurance Policy. Founders are inviting small seed funds into $8 to $9 million rounds for $500K to $1 million as protection against being abandoned. The junior partner who championed you at the big fund leaves, starts their own fund, or moves, and your mandate goes with them. Nobody is left to argue for turning over another card, and the partnership goes back to its real winners. Frankel hates thinking of Founder Collective that way, and says the last 20 deals looked exactly like it. 5. Fewer Entrepreneurs. There are more founders than ever and fewer entrepreneurs. YC professionalized company founding into a norm for people leaving certain colleges, and Frankel thinks many of them lack the fortitude for the learning curve that follows. He watches the trajectories of CEO and CTO split: a good co-founding CTO stays useful in different roles up to 50 people, while the CEO has to learn to manage and to put bums on seats. Mikey Shulman told him 30% to 40% of his time at Suno is recruiting, and Jeff Bezos told a lunch table decades ago that it was 50% of his. 6. CEO-CTO Alchemy. Frankel wants the CTO to be a magician and the CEO to be a good salesperson. In 18 years he has seen that alchemy 4 or 5 times, and it comes from the interplay between 2 specific people rather than from either resume. He prefers co-founders who are different from each other, aligned on direction, and confident in each other's competence. He has said no more often than yes when the second founder looked weak, and he regrets some of those. 7. Ownership Cost Hundreds Of Millions. Founder Collective passed on ElevenLabs, Granola, StarCloud and Fractile purely because the ownership would have been 1% to 2%. Frankel calls it hundreds and hundreds of millions of dollars of lost returns, and it is the mistake he names first. He has never turned a founder down over percentage when he was all in: with Shulman he took the dilution the founder was willing to give, wrote every cent of the first round, and showed the deal to other people. His rule now is that when you meet the right people, you get what you get. 8. I Love It Because. Every Founder Collective team meeting starts with someone completing the sentence "I love it because." If you can't finish it, you can't invest. Valuation comes last in the sequence, after opportunity, market and founders, and Frankel says the price is rarely comfortable. The endings that work sound like insight into a vertical, or an edge nobody else can match in a commoditized business. 9. Vertical Nepo Babies. Frankel is writing a piece on why he loves funding nepo babies, and his definition is founders who grew up inside a vertical. TJ Parker was working in his father's pharmacy at 14, years before PillPack. Evan at Rebar worked for his uncle's HVAC company, one of 10 rolled up by private equity, was told to go find the AI for blueprint quoting, found none existed, and started the company. Over 100,000 US mechanical engineers make at least $100K a year doing that quoting by hand. 10. The Theme Before The Theme. Founder Collective's second fund reads as an applied AI fund today, and applied AI was not a theme when those checks were written. Shield AI was already called Shield AI in 2016. Drones, video cameras and the rest of that hardware are commoditized now, some of it down to $20, and the value came from putting AI around commoditized platforms. Frankel's job is to be in the next theme 5 or 10 years early, and he points out those companies are never the expensive ones. 11. The Last 5%. The SaaS apocalypse is overdone for embedded software and real for everything else. Frankel's test is what runs inside your system: billions of orders in real time, or mission-critical biotech research. Veeva and Olo are much harder to rip out than a horizontal tool a customer can rebuild in Claude over a weekend, which is why he thinks the market threw the baby out with the bathwater. The contrarian trade he does not run would be buying a basket of top SaaS stocks that have lost most of their market cap. 12. Disruption Comes From China. OpenAI and Anthropic will be disrupted, and Frankel thinks there is an excellent chance it comes from China. He is watching the strength of the Chinese open models and notes those labs have not had time to establish incumbency before being taken down. Underneath that, he expects photonic computing to come for Nvidia, because every connectivity piece in a data center that can be fiber already is, and the chip is the last piece that isn't. His broader worry is that the US has cut DARPA and university R&D, which is the money that produced all of this in the first place.