Ask HN: Time management when instructing LLM agent to do work
1 comments
If you are leaving all the work to LLM agents running by some set of prompts, it is too much faith to the machines.
I never give a full permission and always follow the agentic workflow as much as it is being exposed.
I am using these coding or orchestration tools to compensate to fill the void faster under close supervision. Consider them new graduate coder you are mentoring and coaching.
This is the all that I can share about my usage. I hope it helps.
These days LLM agents such as Claude Code/Codex are getting more and more popular, In my daily work, I leverage them to do much portion of the development work, but I am wondering how people manage time efficiently when LLM agent is doing the work.
Take me as a example, when LLM agent doing the work especially I turn on the auto accept mode. I just do the following things:
1. thinking about designs of new task or read some document related of the project I am working on 2. surfing on hacker news:)
but it seems the context switching bother me. when I jump back to check how the LLM agent do the work it seems slowly to get into the context.
Maybe this is the new norm we need to adapt to?