Comment by kkylin
6 days ago
I found this comment especially interesting: "The code in an agent that actually “does stuff” with code is not, itself, AI. This should reassure you. It’s surprisingly simple systems code, wired to ground truth about programming in the same way a Makefile is. You could write an effective coding agent in a weekend. Its strengths would have more to do with how you think about and structure builds and linting and test harnesses than with how advanced o3 or Sonnet have become."
I haven't used coding agents. Can someone who has (or knows about these things) elaborate, or point me to a good resource? Thanks!
It's a while loop. You collect a ton of context (file paths, function signatures, etc.), provide a bunch of functions to the LLM that it can ask to be called and an objective ("Implement X", "figure out X"). Then you send this to the LLM, perform the function calls it requests ("open file", "edit file", "call service X") and append their output to the payload and send it back to the LLM. Rinse and repeat until objective is complete (in which case the LLM may request to call a work_complete function/tool).
That's really it. Stuff the context, loop over the LLM until work is done.
Thank you!