Welcome back, Founders & Operators. Big week ahead: four events in seven days, and an editorial on the open-weights fight that split the AI industry last week.
THIS WEEK AT FOUNDERS & OPERATORS
Rooftop Drinks · Wednesday, 6:30-8:00 PM · Lower East Side: Our flagship event returns. Sips and conversations on GTM, infra, AI tokenomics, and your take on the latest open weights debate.
Final Spots/Waitlist:
[ → RSVP HERE ]
Return of the Quants · Thursday, 3:00 PM ET · Virtual, with The AI Collective
We’ll cover "Building Live Trading Agents: From Prototype to Production." From pulling market data to signals, backtests, and guardrails.
[ → RSVP HERE ]
Build Session · Friday, 1:30-5:30 PM · Harvard Club of New York
Unlimited tea and coffee, deep work, peer feedback, and a tour of the library. One of the most unique rooms in the city to build in. Limited seats.
[ → RSVP HERE ]
Breakfast Club · next Thursday Aug 6 8:30-9:30AM · Chelsea
Our newest format. Coffee, croissants, and curated conversation with founders, operators, and investors, with the option to stay and co-work after.
First edition, small room.
[ → RSVP HERE ]
PARTNER NOTE
Looking for a new banking partner? Rho would like to meet you. They'll pay you $250 for taking a 30-minute call to learn more about their offering. Requirements: You are (1) US-based, (2) not in a regulated industry and (3) have a ~100k liquid company balance.
EDITORIAL: THE OPEN WEIGHTS DEBATE
Last week the AI industry had its most public split yet, and it happened over a PDF. On July 24, Jensen Huang joined X. His first post ever was an open letter on "Open Weights and American AI Leadership," signed by 25 companies including Microsoft, Meta, IBM, Mistral, Hugging Face, a16z, and YC. The argument: open-weight models drive competition, expand access, and keep American AI leadership broad rather than concentrated in a few closed labs.
Within 48 hours the coalition doubled to roughly 50. OpenAI signed quietly, days after the NYT reported it had lobbied Washington to restrict Chinese open-weight models. Google signed Saturday. As of this writing, Anthropic and Amazon are the notable holdouts.
The timing was not random. Three days earlier, Treasury Secretary Bessent warned that Chinese labs conducting "industrial-scale distillation attacks" on American models could face sanctions, naming Kimi, DeepSeek, GLM, and Alibaba. A week before that, Moonshot had released Kimi K3, a Chinese open-weight model that ranks among the most capable in the world, at a fraction of frontier-lab pricing. That release is what set everyone off. So the real fight is not open versus closed. Nearly everyone (including Dario) now claims to support open weights. The fight is whether Chinese open-weight models are a competitive gift or a national-security problem, and who decides.
The case against restrictions: block Chinese open models in the US and you concentrate the market in two or three closed labs, while founders in Europe and Asia keep building on models that cost a fraction as much. American startups would pay more for intelligence than their competitors abroad. A strange way to win a race. The case for caution: distillation of US frontier models is IP theft at industrial scale, watermarks from American models have reportedly surfaced in Chinese systems, and open weights cannot be recalled once released. Washington currently appears to favor case-by-case treatment over a blanket ban.
What it means if you‘re building:
One, model choice just became a procurement question, not only an engineering one. If your stack leans on a Chinese open-weight model, there is now regulatory risk attached. That’s not a reason to avoid them, but a reason to keep a flexible abstraction/orchestration layer so you can swap models in a weekend.
Two, the cost gap is the story underneath the story. Cheap open-weight inference is what makes agentic products economically viable at scale. Last week we modeled the P&L of a voice AI startup; where inference prices settle decides those margins. Whatever Washington does here reshapes the unit economics of every AI product in this community.
Three, watch what companies do, not what they sign. The signatures are positioning and strategic communications. What they actually advocate for in D.C. behind closed doors is more telling.
See you out there,
Stephan