In Claude, with every progress in the step, the code destroys the previous things We created a detailed checklist together in Linear, and at each step as we progress he breaks old things Quite a few Reddit posts point out that this is intentional on your part so that we don’t

🤣🤣🤣

Codex for running overnight. <looks inside> > continue > continue > continue > continue > continue > continue > continue > continue .....

We're giving away an 11-page guide titled: “How to clone your voice and make an agent in <10 minutes” (Like we did for Elon and Karan). It can call support to complain on your behalf, make reservations, prank your friends, etc. You also get $100 in free credits so you can play

Expected completion in 5 years

o3 supports reading 500 lines at once???

Is personal API support with reasoning models no longer available?

When will there be support for right-to-left (RTL) text direction in the chat? It's impossible to use it in Arabic. A conversation about code necessarily includes English terms that reverse the sentence structure, and it's really impossible to discuss it in right-to-left

The main reason we want open-weight models from OpenAI is so that it cannot downgrade them a week after launch Pay attention to the reports: after each release of their model, users always report after a few days that the model responds differently, "as if it's a different model"

The cursor hasn't responded to messages for several hours.