# How Bryan Cantrill actually uses coding agents

DTrace, Oxide co-founder and CTO

**If you take one thing:** `This Old Repo`, reviving a 1990s game. he pointed Claude at BattleTris, a game he started in the mid-1990s that had not built in over 20 years. Claude found the 64-bit stack smash that kept it from running. It is his case for agents doing tedious, open-ended restoration nobody would finish by hand.

- **Agent:** Claude Code (cli) ([source, 2026-05-25](https://bcantrill.dtrace.org/2026/05/25/a-portentous-reunion/))
- **Agent:** ChatGPT ([source, 2026-06-09](https://oxide-and-friends.transistor.fm/episodes/this-old-repo-llms-and-the-restoration-of-battletris/transcript))
- **Model:** Claude Opus 4.8 ([source, 2026-06-01](https://github.com/bcantrill/BattleTris/commit/35e5f85291b1f3b1f8ce7277324401f4177a5e04))
- **Review:** Self-review first: nobody else reviews LLM code until the responsible engineer has ([source, 2026-08-24](https://rfd.shared.oxide.computer/rfd/0576))
- **Harness:** Claude Code on the source tree, kept on a short leash: no closed loop, no one-shotting, and in his first kernel task no builds ([source, 2026-01-15](https://oxide-and-friends.transistor.fm/episodes/engineering-rigor-in-the-llm-age/transcript))

## In their words

> I picked it up, and tasked Claude to port it to Linux.

> And I think I actually went to chat GPT to get its ideas.

> Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

> LLM-generated code should not be reviewed by others if the responsible engineer has not themselves reviewed it.

> I'm definitely not closing the loop, not vibe coding it, not one shotting it, but really because in particular, like, I am not we're not even I'm not even gonna let it build anything.


Archetype (our label): Purist. Newest source: 2026-08-24.
Card: https://www.prehistoriq.com/stacks/bcantrill/ · Data: https://github.com/prehistoriq/stacks/blob/main/data/stacks/bcantrill.json