"Your board doesn't understand the model. Your investors don't understand the architecture. Your team can't know what's keeping you up."
Technical founders at the frontier of AI occupy a peculiar kind of solitude. The problems are genuinely novel. The stakes are real. The decisions — on compute allocation, on safety trade-offs, on who to hire for roles that barely existed two years ago — require a fluency that almost no one around you possesses.
Conferences give you exposure. Advisors give you frameworks. Neither gives you the peer who can read your board deck and tell you what the number actually means.
The loneliness isn't personal. It's structural. You are building something that most of your support system cannot fully contextualize, and that gap compounds every quarter — in your decision-making, in your judgment, in the slow erosion of having no one to think out loud with.
"The hardest part isn't the technical problem. It's having no one who can hold the full picture with you."
— Infrastructure founder, Series A
Twelve seats.
Four gatherings.
Zero spectators.
The structure is minimal by design. What fills it is the conversation that cannot happen anywhere else.
Seats
The table is capped at twelve. Not because of the venue — because trust requires specificity. You will know every name. Every company. Every burden.
Gatherings
Quarterly. Two days each. One city, chosen for quiet: a ranch outside Marfa, a house in the Sonoran desert, a property in the Santa Ynez hills. No conference center. No badge lanyards.
Spectators
No press. No investors in the room. No one whose livelihood depends on what you decide. The Chatham House rule governs everything said. What is shared there stays there — fully.
Agenda per session
Each gathering has one shared agenda item submitted in advance — a real problem, not a polished case study. The rest is unstructured time. The hallway is the point.
Membership is not purchased. It is offered — after a conversation, after review, after we are certain the table is better with you at it.
Current cohort
Applications open for the Q3 2026 cohort.
3 seats remaining.
Three entries from the room.
Names withheld. Companies obscured. The problems are real.
Foundation Model Co-Founder
Pre-Series B · 18 months post-launch
I had been carrying a decision about our inference architecture for eleven weeks. I had modeled it six ways. I had talked to advisors. I had read every paper. I sat down at the table, described the problem in full — the real version, not the board-meeting version — and within forty minutes I had a perspective I had never considered. Not because anyone was smarter than me. Because they had all made a version of that decision, and I hadn't.
Applied AI CEO, Enterprise Sales
Series A · Fortune 500 deployments
The thing no one tells you about enterprise AI sales cycles is how isolating the legal phase is. Your lawyers are protecting you. Your investors want the logo. Your team is waiting on the contract. And you are alone in a room with a procurement team that has never bought anything like what you are selling. I brought that exact problem to the table. Three of the twelve had been in that room before. The conversation that followed changed how I run every deal.
GPU Infrastructure Builder
Seed · 7-person team
I almost made a $4M commitment on compute that would have been wrong. Not obviously wrong — the kind of wrong you only see in retrospect, after you've burned six months and your roadmap has quietly shifted around a constraint you locked in too early. Someone at the table had made exactly that mistake. They didn't tell me what to do. They told me what question I hadn't asked. That was worth more than any advisor I've ever paid.
The question you are carrying alone right now — someone at this table has carried it too.
Apply for a seat.
The form is short. The last question is the one that matters. No ticket price is shown at this stage — membership is offered after a conversation, not a transaction.