AI Treaties, Nathan Fielder, Agents (of Hollywood), and Grill Evals

MORE OR LESS — Agentic Commerce Discussion (obviously) — topic of the week… we get into how this will all play out on business and policy terms (now that the tech of give everyone a CPU in the cloud) is feeling solved…. Amazon vs. Meta, my inability to book the Hilton, etc. are all becoming the question (fast)… as they always would. We develop the key idea live on the show of the ATF — Agent Treaty Framework…. Because standards are STUPID in the era of AI, and as the world de-globalizes in every way, it is all about treaties. I also gush about Nathan Fielder & brit starts her peptides.

THE IMPORTANT STUFF THIS WEEK

• NATHAN FIELDER - You can see everything - I flew down to LA last min for the premiere there… It is beyond. It might be the most amazing movie I have ever seen. Makes a legit intellectual / emotional contribution, is completely insane, wild. I think the reason details haven’t leaked as more people have seen it is a sign of how much respect everyone who has seen it has for the film and art. Also, total justification of the 3 hour movie as a format (it couldn’t be a short, it couldn’t be a series). I love comedians, and have always been a Nathan Fielder fan… but this is next level brilliance.


• AI STANDARDS ARE STUPID, AI TREATIES ARE THE FUTURE. Key point… MCP was stupid, this latest ‘commerce’ standard is even dumber. Standards come when a hegemon forces them OR when all that matters is we drive on the same side of the road… Agent access to commerce, etc. is war and pairwise. The next era isn’t about tech, it is all about business and politics… And here is the key thing, AI makes treaties possible not just for big companies, but all the way down… the beauty of ‘have your AI negotiate with my AI for access’… is that companies can manage thousands / millions of deals at a time (vs. 12 via human biz dev)… so the future is a web of negotiated truces … and it isn’t an open one.


• (HOLLYWOOD) AGENTS THE FUTURE OF AI AGENTS... which leads me to Ari Gold (I mean Ari Emanuel)… who knows, great book. The funniest part in all of this is that AI agents are going to work hilariously just like old school Hollywood. There is no ‘standard’ API… my agent calls your agent, you get a call back based on how powerful your agent is (who is powerful if they represent the best talent / most valuable customers)… An API isn’t a ‘key’… it is a call ‘hey I want this’ response ‘well tell me about your customer / client’ … and then a pitch back - and if you have a great agent with a great trusted firm / client roster — you get better performance and engagement (if you are some unknown good luck)… so what? So just like in Hollywood, the AGENTS that survive will have the best most valuable customers/clients, and will have to be big. The future of all of this is hilariously plain in simple in the past — the tech gets out of the way, and agenting is just agenting… back and forth negotiations for every deal / request (because you can), and leverage based on who your clients are, and which firm is repping you.


• EVALS ON SMOKING MEATS — hilariously one of the things I am better known for is being in the background of the ‘Zuckerberg meat smoking remix videos’ — 10 years later, it was fun to watch the reprise (we old)… and get into the anti-trust issues of vertically integrating muse with the pit viper for smoking meats, and the obvious need for real world evals to judge how good each agentic platform is at cooking the perfect brisket


• AI LIABILITY FRAMEWORKS — per last week’s screed (and lots of shouting before that), and good to see the NYT catching up on the fact that practical liability around AI labs is the whole game. There will be lab leaks (just like wuhan)… but we can’t make US tax payers pay for the externalized risks that the labs are knowingly creating. They need to foot that bill themselves either with insurance coverage (where they will be financially incentivized to fix their controls because they won’t be able to afford the reinsurance otherwise) OR through paying into a federal bailout program for the inevitable moment when FEMA is going to be called in to clean up their mess… When it is this important it is criminal to let them internalize the upside, and radiate the risk to the American taxpayers.

OTHER HOT TAKES

• READING ALL THE HEDGE FUND QUARTERLY LETTERS — for the first time i can remember i'm genuinely interested in how they're thinking about the moment… some are quite good. the ones up 30% on the quarter do like talking about small businesses.


• "QUICK SNAG, HIT 'VERIFY YOU'RE HUMAN'… WANT ME TO CLICK IT?" — state of the internet 🧑‍🍳😘 … my agent politely asking permission to lie about being a person is the whole 2026 economy in one screenshot.


• STILL NO MOAT IN AI AGENTS — i said i'd die on this hill and 176k impressions later nobody has handed me the counter-argument… i switch between muse, instinct and dots at basically zero cost — one github repo, a few databases, feed the new thing my context, done


• SKIP-LEVELS FOR THE BOT ARMY — i already have a bot chief of staff managing the army… at what point do i owe my workers skip-levels, Q&As, and office hours?


• AI COLD EMAILS ARE AT LEAST FLATTERING — a little hit of ego every time you open something from an unknown sender these days… the pitches have never been this well-researched.


• YOU AREN'T SELLING OUT… YOU ARE SELLING UP — free rebrand for everyone taking the big check this cycle.


• ELIZABETH HOLMES QUESTION — if everyone had invested by handing her bags of cash rather than wire transfers, would she be a free woman today?

Best,
Sam

P.S. "After seven and a half years living in America, i think i know the vibe" (Maverick, 7 and a half years old)