Just whipped up a programming language MVP with the Claude app. Just for fun ๐
The main idea is that everything is function-call style, like +(1, 2).
@ElimDroflem Thanks, you were right. OpenAIโs status page confirms an ongoing ChatGPT conversation incident. I found that sending a simple follow-up prompt usually resumes the same Codex task, so starting a fresh task isnโt always necessary.
Codex (gpt-5.6-sol xhigh) explained to me when to use xhigh and high:
A good split is:
- xhigh: planning, schema/API design, security review, workflow implementation, and difficult cross-layer debugging.
- high: UI polishing, focused fixes, tests, documentation, and routine follow-up work.