If we want models to behave well in the real world, we need to train and evaluate them in something closer to the reality. Releasing carla-env, an open source embodied environment that gives models access to engineering scale physics simulation. All details are in blogpost:

I see a future where people build custom software with embedded coding agents that perpetually keep it up to date and optimised

rl is such an interesting field to me considering i'm someone that used to struggle with learning(still do)

Introducing the💡 Ideate agent in Stitch. Use it at the start of a project to explore what’s possible before you converge on a solution. It takes a moment to think deeply, fetching context and exploring ideas so you don't have to. Think of it as your design buddy that: 🌐

Deployments are live train a model, then immediately use it for inference

you can launch cloud instances from codex cli

i could read codex's cot all day

github codex reviews has really been eating into my usage 😩

these twitter bots are getting out of hand

new codex release is here with multi agent improvements

I basically need a codex subscription for each repo 💀i'm cooked after the 2x rate limit period is over

are you really koding if your skills don't have weird names?

5.3-codex cot is so interesting. It's always encouraging and hyping itself up

there is so much cool stuff being created everyday! I can't keep up 😭

shouldn't codex 5h rate limit technically prevent you from using all of your weekly limit before the week ends?

we need a good open source GUI for these CLI tools

agi has probably been achieved, it's just a systems and infrastructure issue now

don't write tests, simulate, simulate, simulate

5.3 in a review loop = near perfect software

finally we'll be able to focus into codex agent threads