An interesting question is whether the future of AI will look a lot like the present in Vertical SaaS. The models builders are horizontal, but the actual enterprise applications most companies will use will still be industry specific (like this one).

Overheard: "The big beautiful bill shows US Government is the most AGI-pilled. They know GDP is going to double in a few years so no need to worry about the debt!"

Asset protection from human drivers is underrated

People like to talk about how AIs are better at evaluating than generating. But so are humans! It's way easier to critique an essay, song, or movie than it is to make a better one.

Center left twitter is being played like a fiddle by Zohran in the last few weeks.

As companies in regulated industries evaluate how to use AI, they are running the "in-front-of-Congress test," says Valon CEO Andrew Wang. (Essentially, if something goes wrong, how will they explain the decision to use AI to the government.) "This is a huge fear factor for

Any “abundance” org that doesn’t rank Cuomo over Mamdani should not be taken seriously. Totally insane

An underrated problem is that, due to (seemingly?) legal concerns, modern ATS software refuses to let you rank candidates using AI or any other means. The fact that you can't makes companies not even look through the pile of thousands of resumes, and both sides are worse off.

Is this the largest burn for a private technology company in history? o3 says Meta burned 17bn in 2024 on Reality Labs. Uber was 5bn at its peak, WeWork 2bn. OpenAI lost $5bn last year.

Using o3 feels like Gell-Mann Amnesia on steroids. Most results look great, but then it takes one result on something you know to make you question everything.

Obviously Anthropic is focused on enterprise but still always shocked by how big the gap is between them and ChatGPT