President Trump says that Sundar Pichai talked him out of renaming the Atlantic Ocean. 'And I spoke to a brilliant guy, Sundar, he heads all of Google, he's a brilliant guy. Sergey and Sundar, totally brilliant guys. You know, they're still rattled from the Gulf. And I said, "Sundar, I'm thinking of the Atlantic Ocean, can you handle it?" "No, no, no, this time I’m not doing it," he said. He said, "Oceans are different, sir. You know, they cover, they cover really a lot of countries. Please don’t ask us to do that."

watch mcdaniels as the shot go up bro is funny asf he emote every time somebody take a 3😂😂 bro put a cigarette out and stepped on it

I waited, learned and made my first investment since returning as an active investor at Sequoia. Whop is a special company, and I’m excited to help them to create jobs and stimulate the economy.

the silliest way to lose a lot of money in 2026 is to hold onto stupid priors for dear life the second silliest is to make “safe” bets behind prestige vs. first principles evaluation of tech/business. if a company has fancy pedigree but is just losing…don’t fight the tape

base:0xed664536023d8e4b1640c394777d34abaff1df8f served ~32bn DeepSeek API tokens through its inference network over the last 24hrs - around 6% of the ~500bn/day Venice reported last week - whilst the network remains in effectively closed beta.... Usage is still predominantly running on promotional credits so not a revenue signal (yet), but the throughput confirms the network is capable of production-scale volume. 💪

Thanks ChatGPT! You’re absolutely…not right. When I was saying check your numbers — I didn’t mean total training cluster build commitments, I meant the 95% GRR, 400% NDR of the inference leaders. That is best-in-class stickiness and close partnership with every significant customer, not “reverse auction with a one line api change.” What does a Neocloud model that builds and sells its compute in long term commitments, underwritten against IG credit, get you in the inference business? It doesn’t get you the serving stack, the production muscle, the customer trust, or the freely allocated supply. Inference is a huge market, and it’s likely that new companies will find new niches. Like Digital Ocean after AWS. But in a supply constrained market you’re betting somehow the new players find a better use of the supply than the existing players. What does a new player at the bottom have as a business advantage? These businesses need to maximize dollars per MW in order to even afford the compute for their customers. Unless they’ve got strategic friends, where are they going to get their supply? If they can only peel off the most cost sensitive, lowest margin workloads as a cold start (say, batch?) and can’t fit it into an overall cloud’s utilization picture, how can they afford market supply? Without the engine to identify and underwrite new suppliers in a sea of scammers, how can they even find it? If your customers grow…you quickly need a gigawatt. People only talk about megawatts because they can’t afford it. In inference, I’d attack supply and not demand — I would be most excited to hear about a new player that has a clever idea of where they, uniquely, are going to get a gigawatt. VC funding is the silliest indicator of startup success. It is also, by dollars, going to get drowned out by the funding the inference players raise. Hundred of millions is small ball. You think the leaders are serving generic weights? It’s 90% custom, 1000s of different models, and any given big customer uses a constellation of models (often tens) in their product. Post training (and specifically training and tuning against our stack) is an engine that produces more value for our customers, that they reinvest the savings in customer growth, better product and in inference on us. Customers are wonderfully hungry, and so a good inference company will rapidly deliver them more: sandboxes, ecosystems and better abstractions. This is not a liquid commodity game, it’s a platform one, with a supply prerequisite. Anyone big who splits traffic only does amongst a small number of leaders, with custom models. And they do not churn — unless you can’t keep up with their growth, or fuck up reliability or security.

*OPENAI EYES ACQUISITION OF CRYPTO SECURITY COMPANY TRAIL OF BITS: THE INFORMATION

Ladies and gentlemen, we are preparing for landing. Please fasten your seat belts, return your seat backs to the upright position, stow your tray tables and open your window shades. Thank you.

Startups selling data to AI labs are putting up unprecedented sales growth but it makes sense. In the stock market semis selling memory/compute to AI labs are the best growth stories and AI improvement is basically compute + data.