SpaceX's GrokBot team shipped 11,000 PRs in a month — Lauren Tan explains how
FinanceYF5 · x · 2026-11-13
Key facts
Lauren Tan (poteto), formerly React Compiler core team, now principal engineer on SpaceX Superintelligence's GrokBot, revealed at AIE CODE 2026 that the team shipped 11,000 PRs to production in October, with most product code no longer needing deep human review; rollback rate fell from 0.98% in June to 0.47%, while app startup, typing and session switching got faster than May.
The three-part playbook
- Fix the codebase first: outright ban patterns agents get wrong; directory structure is context; comments cut from 66,000 to 1,200.
- Review becomes watching verification: humans don't read diffs — they watch screen recordings, screenshots, state diagrams and formal verification output; most tokens go to verification.
- Don't automate a bad loop: if the inner loop produces slop, automation just scales slop; build the outer loop (auto-find bugs, auto-fix, auto-merge) only after the foundation is solid.
The core rule
Every time you correct your agent, turn that correction into a deterministic check in the code (a lint or categorical elimination of the problem class).
GrokBot itself
An Electron desktop app launched Aug 11; it's both the team's daily coding agent UI and the "outer loop," ingesting X feedback, Linear tickets, Slack messages, cron jobs and Datadog/Sentry alerts to dispatch agents automatically.
Related event: SpaceX engineer shares seven-step playbook for trustworthy coding agents(6 posts)→
More from coding & agent
- Gemini /teamwork agents caught live: four parallel task logs all running that second — RyanSpunk · 2026-11-13
- Codex CLI 0.168.0 adds model provider picker, side-by-side TUI workspace and animated GIF generation — github-actions[bot] · 2026-11-13
- Free MCP server lets Claude read Altium, Allegro, KiCad and Gerber PCB files directly — woodisruff · 2026-11-13
- a16z Partner Clones Open-Source Game with Fable 5.2 Agent, Which Finds Its Own Horse Animation Asset — FinanceYF5 · 2026-11-13
- OpenAI team shares how it runs and governs fleets of persistent agents internally — romainhuet · 2026-11-13
- Perplexity's Decider scanned 1B tokens across 7,000 agent sessions, found 14 security fixes — denisyarats · 2026-11-13