An Ex-Meta L8’s Agentic Engineering Setup
A former Meta L8 engineer shares his end-to-end agentic workflow, from terminal setup to prompt strategies. He outlines the tools (WezTerm, tmux, Neovim, Claude Code/OpenCode), voice input, and delegation techniques that let him manage AI agents like a dev team.
- A former Meta L8 produces 30+ high-quality PRs in a single day by managing AI agents instead of writing code himself
- His entire workflow runs in a terminal (WezTerm + tmux + Neovim) using Claude Code for Anthropic models and OpenCode for everything else, avoiding vendor lock-in features like auto-memory
- He pairs local, offline speech recognition with terminal prompts to iterate faster, treating each prompt like a spec refined against a "golden set" of expected outputs
- He's built custom CLI tools to automate context loading, versioning, and error handling so prompts plug into a repeatable pipeline
Kun Chen stopped writing most of his own code and now treats agents like a dev team. He stays at the top level—deciding what to build and judging quality—while a suite of open-source tools handles the heavy lifting. His example project, “Hi Bit,” is an AI tutor for his son. In a single slow day he produces over 30 high-quality PRs. That throughput comes from refining real-world workflows, not chasing demos or hype.
He works entirely in a terminal. WezTerm runs as one frameless window. Tmux splits it into panes and tabs for agents, code, logs and tests. Neovim—with oil.nvim for file browsing, neogit for diffs and snacks.nvim for search—lets him stay keyboard-driven. For agents he uses Claude Code with Anthropic’s models and OpenCode for everything else. He avoids gimmicks like auto-memory so he can swap in any model from any vendor without rewriting prompts.
Prompt writing itself is a key lever. Chen pairs terminal prompts with local speech recognition—fast, offline, and free—to speed up iterations. He treats each prompt like a mini spec, iterating until responses match his golden set. Over time he’s built CLI tools that automate context loading, versioning and error handling, so his prompts plug into a consistent pipeline every time. That end-to-end setup keeps his hands on the keyboard and his mind on design, not busywork.