So many years reading his newsletter, and forever grateful for Om. He replied to a cold email a long time ago with very specific, helpful advice on a project I was working on. What a legend!
I was surprised today by how much better GLM-5.2 was than GPT-5.5 at aesthetic/UI work. I'll keep my Claude/Codex setup via Conductor for now, but this model got me to set up OpenCode, download their desktop app and do most of my work there today.
The interesting part here is not whether Anthropic is right on safety, but that safety gives them a moral vocab for bold policy changes and platform power.
Yeah, this is similar to finding 10 customers who love your software rather than 1,000 who only kinda like it. And if you do, it's probably because you've found a use case that is urgent enough for one specific kind of user.
Humm maybe. But a plain model sampling outputs obviously isn't doing discovery in the AlphaGo sense. But once you put the model in a loop with tests, feedback, tools or even a human picking the good result, it starts to get much closer to the process he's describing.