listen i love codex more than some of my extended family members but man oh man some of you are getting way too comfortable having it do absolutely everything.
everyone can and should use codex. part of what makes these models so useful is the enormous range of human thought, experience, and expression they’ve learned from. all those different voices, opinions, ways of seeing things. we certainly need to keep adding to that.
you’ve spent your whole life developing your own voice, opinions, feelings, biases, passions. all that shit. your combination is unlike anyone else's. people want to see that. sometimes you say something badly or mess up or change your mind halfway through explaining it. and you look like an idiot, or your tone is off, or maybe you sound like a real asshole. and then you apologise. show people those parts too, when appropriate. it humanizes you and can bring you closer to the people around you. and humor. man, we fucking need humor. fixes everything.
anyway, enough of being soft on the internet. all i’m saying is, use these amazing tools. just know when people want YOU.
We’re opening a research preview of Claude Duet, a new app where Claude works as a proactive agent.
It comes to you with what needs your attention, completes tasks you ask of it, keeps working when you step away, and checks with you before it acts on your behalf.
Ran tsrs vs Go TS 7.0.2 on an app I’m working on, 4 checkers each.
- tsrs: 60ms, 169 MiB
- Go TS 7.0.2: 260ms, 520 MiB
- Go TS 7.1 nightly: 210ms, 496 MiB
About 4.3x faster with 68% less memory than 7.0.2.
Introducing tsrs: a Rust port of the TypeScript compiler 🦀
We are not forking TypeScript!
◆ In close collaborating with the TypeScript team
◆ Up to 8x faster than TS 7.0 & ~70% less memory
◆ Runs in WASM, so it works in your browser & on Workers
◆ Soon part of the Oxc toolchain
npx tsrs
https://t.co/qmmxF3YnQy
If you patched for the eMudhra CA compromise, check the env that's actually running, not just your lockfile.
One of my repos pinned the patched certifi. Its live Python env was still on the old one. Run pip show certifi inside the real env.
Day 16/
You can now cache for 12h in the API and for logged in users across Codex & Work we keep cache forever and you don’t have any penalty for coming back to an old conversation or resuming after a few hours and are now guaranteed to hit the cache.
Fixing bugs with my Dot that I couldn't get past before. I describe the problem, it splits the work across worker threads and comes back with the fix landed. I never have to talk to the workers directly.