Skills, not agents
I used to think shipping an agent product meant building:
- The core LLM loop.
- Skills: describing to the LLM how to do particular things (including building API connectors etc).
But it seems that Claude Code and Codex are strong enough general-purpose agents that you can just outsource (1) to one of those generic ones, and focus just on the skills.
I am trying to take this seriously. For now, in side projects: some time ago I had an idea of a chat-based tool that helps you understand (Estonian) goverement data: statistical reports, open datasets, etc. But building the entire thing is a lot of work, from frontend to backend to datasets etc.
If I can now do it by just describing the data sources and maybe providing a couple API keys, and plugging those into Claude Code, that feels 100x easier. So I'm trying.
Stay tuned.