@morganlinton I hit the same pattern on a long rewrite task (~40 tickets). I ended up running a 3-minute loop to keep pushing it through the rewrites.
Each turn took around 50s to 2min20s max
@badlogicgames I thought you'll put a Pi tile on the front ahah. What are the specs ? It's pretty small which is nice to not take the entire desk space
@Capetlevrai Ahah, il y a pas une réponse identique dans le thread. Blague à part, time toi une heure, test le et vois si t'aimes bien. Au mieux t'auras trouvé ton nouveau framework de prédilection au pire t'auras appris un peu comment laravel fonctionne.
Codex Computer Use, but with a tiny Eurotrip detour 🇪🇺🇬🇧
I packaged a simple MCP wrapper to hook Codex into the locally installed macOS Computer Use, even when the border says “not today” :)
https://t.co/0qOec0na7r
@_smontlouis Depuis quelques temps j'utilise une logique Plan -> Plan review -> Implementation https://t.co/8RyfMY8sc5
Ça fait vraiment la différence et je perds moins le "fil"
It feels like everyone is racing to ship first flooding the feed with new AI features while compromising on quality. LLM make code cheaper to produce but shipping a feature is not just about writing code it is an end to end process that still needs safeguards, including proper QA. Honestly last week i caught myself thinking that i had been spending more time on QA than on coding lately, partly because of LLm that Generate code