On programming with Agents (from Zed blog)
Yesterday I was quite interested in an article by Zed about working with agents. I work with them in a very similar way. The main idea: LLMs automate writing, not thinking. The three rules that are described there: Use the agent only for tasks that you can already do yourself