Eric S. Raymond shares his AI coding workflow: Threadripper rig, Codex, no autonomous commits

esrtweet · x · 2026-11-10

Eric S. Raymond details his setup and deliberately minimal AI workflow:

Hardware & environment

64-core Ryzen Threadripper 9978X, 128 GB RAM, 3.84 TB storage, Radeon AI Pro R9700.
openSUSE Tumbleweed (rolling release with image reversion, Apt/dnf-style dependency handling); tiling WM, recently switched from i3 to Cosmic Desktop.
Three 4K monitors: portrait center for Emacs (116 readable lines), right for Brave, left with three terminals (middle running Codex). Unicomp Model M keyboard, Handy for speech-to-text.

AI usage

Codex with the current OpenAI frontier model; Qwen 3.8 running locally via LLM Studio as fallback.
Keeps AI use deliberately simple: skills, subagents and "software factories" are mostly pointless scaffolding around models that will be replaced in weeks or months.
Codex may read and write files but never commits on its own — he reviews every diff, splits work into small tasks, and keeps firm architectural control, nearly the opposite of vibecoding.

More from coding & agent