I tried it out, it built a system, calling itself AGOS, created a test project and coded it up successfully, added validation checks, verified everything worked properly, made a bunch of scripts to track goals and tasks and other stuff.

I did switch it to use ZODB for database instead of postgres, and fossil-scm instead of git, which it was able to successfully navigate. I'm a python head so this makes sense for me, ymmv - I have 30 years of development experience and that's what I choose to implement.

The hot/warm/cold didn't seem to materialize though, but the task focus kind of makes that a moot point.

I tried with claude, qwen, and gemini, all seemed to pick up and continue working. Work seems kind of slow, but it was all reviewable, clear goals and tasks, and overall kind of usable? I'll experiment a bit, but it seems promising. I still don't know why you wouldn't just have it build you a system then release that.

I don't understand this.

If it's a simple, extremely long prompt that should be copied and pasted without customisation, wouldn't it make more sense to build it as a set of skills?

this prompt is extremely wordy, contains a lot og general considerations that I don't think the llm will take as actionable, repeats itself at a shallow level of abstraction, and is overly broad in its envisioned scope. one does wonder what the person who painstakingly typed in all of this text must have been thinking. also, where are the comparative stats about its prowess on representative tasks?

There goes all your token budget…

That’s a very big prompt