It's an easy way to compare the coding and creative strengths of models. I prefer them over reading a tabular comparison of benchmarks which you have no real insights into.
True, but resetting context to avoid bias doesn't require different agent personas with limited scope, it requires refreshing the context which can be done with a single agent.
I got it to transcribe this: "Create tests and ensure all tests pass" and instead of transcribing exactly what I said it outputs nonsense around "I am a large language model and I cannot create and execute tests".
I've kept the codex-container.sh version going for this eventuality.
Those scripts are for running the docker command with all the ENV vars and settings. Whatever does that does NOT have to be Powershell if you don't want it to be.
Poweshell is awesome because no external tools are needed to modify dates, or do math, or generate JSON or XML, both of which are first-class citizens. Also, in powershell things are objects not strings, so passing something along with a | is much more powerful. Still, if someone doesn't like something, there's always bash!
