When I first saw the title of OpenAI鈥檚 blog, I wasn鈥檛 too excited. Here it is: Building self-improving tax agents with Codex. Tax agents don鈥檛 sound exciting. Everything is self-improving. I got it. But this short post is a masterclass in practical, hands-on learnings on the latest tools, workflows, and approach to how to do evals for real-world use cases. It鈥檚 not just technical; it builds intuition by relying on real-world experience. ...
Open source is back. RSS is back. AI agents make this possible, but only if you make your repo AI-first. Humans still decide what to do. One of my open source repos is becoming the de facto way people convert blogs into RSS. As far as I know, it powers most Anthropic RSS newsfeeds. I鈥檓 getting a surge of new contributors. Agents make coding easier. That part is obvious. What is less obvious is this: they make contributing easier. ...
I had a multi-hour conversation with a single Gemini session and noticed that the responses dropped in quality and started coming back very fast. I鈥檓 familiar with context limits and context poisoning, but the sudden brevity still surprised me. I asked the agent what was going on and found this: It turns out that deep into a work session, 3-line responses are a feature, not a bug. Nothing too deep here, just a fun way to accidentally stumble onto a technical detail through a random observation.