I wonder why they're trying to establish separate branding for hardware. Considering that OpenAI's strongest advantage right now is the ChatGPT brand and they're anyway cutting efforts on other products, wouldn't it make more sense to use the ChatGPT brand?

They certainly don't seem to have a problem with using the same name repeatedly given the 300-or-so products called Codex at OpenAI.

This seems to suggest model releases are going te become smaller and faster in reponse to increased competition, which will be interesting.

On the one hand, labs may finally be forced to focus on making faster and smarter models and becoming profitable instead of playing with experiments, which other companies seem to be better at anyway.

On the other hand, this could make the endless torrent of AI news even more pervasive, and more importantly could make labs focus on short-term gains instead of larger breakthroughs.

A lot of these kind of bugs feel like they could be caught be a simple review bot like Greptile... I wonder if Cloudlare uses an equivalent tool internally?

What makes greptile a better choice compared to claude code or codex, in your opinion?

That has not been my experience with those tools.

Super-procedural code in particular is too complex for humans to follow, much less AI.

Any bot that runs an AI model should not be called "simple".