"So GPT-5-Codex is perfect?
Absolutely not. Sometimes it refactors for half an hour and then panics and reverts everything, and you need to re-run and soothen it like a child to tell it that it has enough time. Sometimes it forgets that it can do bash commands and it requires some encouragement. Sometimes it replies in russian or korean. Sometimes the monster slips and sends raw thinking to bash. But overall these are quite rare and it’s just so insanely good in almost everything else that I can look past these flaws. Humans aren’t perfect either.
My biggest annoyance with codex is that it “loses” lines, so scrolling up quickly makes parts of the text disappear. I really hope this is on top of OpenAI’s bug roster, as it’s the main reason I sometimes have to slow down, so messages don’t disappear.
Conclusion
Don’t waste your time on stuff like RAG, subagents, Agents 2.0 or other things that are mostly just charade. Just talk to it. Play with it. Develop intuition. The more you work with agents, the better your results will be.
Simon Willison’s article makes an excellent point - many of the skills needed to manage agents are similar to what you need when managing engineers - almost all of these are characteristics of senior software engineers.
And yes, writing good software is still hard. Just because I don’t write the code anymore doesn’t mean I don’t think hard about architecture, system design, dependencies, features or how to delight users. Using AI simply means that expectations what to ship went up."
https://steipete.me/posts/just-talk-to-it
#AI #GenerativeAI #Codex #LLMs #Chatbots #SoftwareDevelopment #Programming #AgenticEngineering