# How Salvatore Sanfilippo (antirez) actually uses coding agents

Redis creator, DwarfStar (ds4)

**If you take one thing:** `QA_BEFORE_RELEASES.md`, an agent QA pass per release. a 2,088-line release gate an agent works through, which he credits with raising quality.

- **Agent:** Codex (cli) ([source, 2026-05-04](https://antirez.com/news/164))
- **Agent:** Claude Code (cli) ([source, 2026-02-24](https://antirez.com/news/160))
- **Agent:** ds4-agent (cli) ([source, 2026-05-19](https://antirez.com/news/166))
- **Model:** GPT-6 Astra / GPT-6.1 Sol ([source, 2026-10-03](https://x.com/antirez/status/2106312728629387347))
- **Model:** Fable ([source, 2026-07-13](https://antirez.com/news/169))
- **Review:** Still reviews Redis code, calls it mostly pointless ([source, 2026-07-13](https://antirez.com/news/169))
- **Harness:** tmux / Terminal.app, driven from Telegram via tgterm ([source, 2026-09-07](https://github.com/antirez/tgterm/blob/51433ee538d740b013b117e4e5cbcebbddf86bc2/README.md#L9))
- **Rules file:** AGENT.md per repo, CLAUDE.md symlinked to it ([source, 2026-06-05](https://github.com/antirez/ds4/blob/297f750cda347d3a7b6012564ec8538fc4712a34/AGENT.md#L18))

## In their words

> I started writing for days a long specification by hand, then I paired with Opus initially, then GPT 5.3 was released and I switched all the design and development with Codex. Since then I use only GPT 5.x for system programming tasks.

> I thought it was time to try a similar experiment myself, one that would take one or two hours at max, and that was compatible with my Claude Code Max plan: I decided to write a Z80 emulator, and then a ZX Spectrum emulator (and even more, a CP/M emulator, see later) in a condition that I believe makes a more sense as “clean room” setup.

> I have the feeling I can decide which is better only with enough practical evidence, by using ds4-agent over multiple sessions with the two systems.

> Having serious issues to work at Redis with Astra / Sol 6.1 because of continuous firing of cyber checks.

> Fable and GPT 5.6 reviews to the sorted sets memory saving are going to spot ways more errors and subtle race conditions that my review is going to uncover.


Archetype (our label): Toolsmith. Newest source: 2026-10-03.
Card: https://www.prehistoriq.com/stacks/antirez/ · Data: https://github.com/prehistoriq/stacks/blob/main/data/stacks/antirez.json