The post does explain in quite a lot of detail, the user was running `claude` or [Claude Code](
https://docs.anthropic.com/en/docs/claude-code/overview) an Agentic editor that runs in the terminal. You can run it within a project and provide prompts and context to it, the LLM will read through these things, perform searches if it sees necessary (and you allow it) then it will prompt you with edits or tool usages (cli commands or file edits) that you can approve, deny or suggest edits/changes.