The BugBash Podcast From Scale to Rigor: An Engineering Journey at Meta and Oxide
Mar 11, 2026
A journey from massive, data-driven engineering to shipping untouchable, air-gapped hardware. How technical writing became a force multiplier for complex designs. The surprising power of property-based testing and Oracles to catch bugs early. Practical use of LLMs for prototyping and doc review with strong human oversight. Real-world challenges of testing and operating a cloud-in-a-box without live patches.
AI Snips
Chapters
Transcript
Episode notes
Start Writing By Naming Your Audience
- Before writing, name your audience and step into their shoes; get reviews from people in that audience and use LLMs like Claude for perspective checks.
- Calibrate detail level: source-control team docs assume background knowledge while public RFDs require broader explanations.
Air Gaps Force Postmortem And Resilient Design
- Air-gapped, customer-run hardware forces heavy upfront thinking because live monitoring and remote debugging are unavailable.
- Oxide uses support bundles, post-mortem analysis, and resilient degraded-mode behavior to compensate for zero internet observability.
Pass Seeds And Use Property Tests For Reproducibility
- Use property-based testing wherever you can express invariants or stateful sequences and invert non-determinism by passing deterministic seeds.
- Seed injection and RNG expansion make tests reproducible across generated histories and help find inner-loop bugs early.