推文

I really like this technique that I learned from @nbaschez , where you keep shipping features all day and just merge it to a next branch. And then at the end of the day, you just check if everything looks good in the next branch and ship that to main. It's too much work, too much of a burden to really in-depth review 20 pull requests per day, but it's very doable to have agents merge all these pull requests automatically into this branch and then test everything at once and merge.

wow, more tokens spoken into @cursor_ai than claude + codex combined last month! There are two ways I work with AI now. One is in a loop, talking to it using @usemonologue in @cursor_ai , and the other one is in the background, orchestrated, running overnight and long. Using tmux and @claudeai code cli and @OpenAI codex cli. I do both more than I did before.

After using @cursor_ai 3.0 for 5 days, it really starts to grow on me. It's not there yet, but it's getting there and I see a lot of potential, super excited for what the team is cooking on. Try it out, there is something happening here. https://t.co/OHIaz0HCba