ErnieFull-stack engineerTypeScript, React, Ruby on Rails
I find where AI coding agents break, and build the checks that catch it.
- Paper
- Measuring the Wrong Number · peer-reviewed, ACM 2026
Blog
Where AI autonomy breaks2026·06·05Passed every test. Still took down prod.It passed every test and still took down prod — the bug wasn't a bad line, it was a state nobody enumerated.2026·06·05The invoice was dead. Then it took your money.A voided invoice that still captured a payment — green tests, red ledger, and a bug that lived in the ordering, not in any single line.
All writing →
Cut AI costs
Per token or per seat, I cut what your coding agents spend: a smaller API bill, or the $100 plan instead of the $200 one.
Free 30-minute bill review →Background
- Earlier · Full-stack
- Shipped products end to end for North American startups — React and TypeScript over Rails.
- 2022–26 · archive.com
- Built the Claude Code and Codex harness the whole engineering team works through, and LLM features with evals behind them.
- 2026 · Research
- Agents on a real codebase raised questions nobody had measured. Measuring them became the work.
- Open source
- vigilesChecks that a coding agent's rules (CLAUDE.md, skills, hooks) match the real repo: every rule it cites exists and is on.
- Open source
- paperlintCatches what gets a research paper sent back: page limits per venue, a silently swapped font.
- Sep 2026 · Peer review
- Reviewer, Agents in the Wild workshop at NeurIPS 2026, 2 papersNeurIPS is the largest machine-learning research conference. This workshop, on keeping AI agents safe and secure, has also run at ICLR and ICML.
- Aug 2026 · Peer review
- Reviewer, REALM workshop at EMNLP 2026, 1 paperEMNLP is one of the top conferences in natural-language processing. REALM is its workshop on research into language-model agents.
Appendix — the full stack, if you're curious
ShipReact · TypeScript · Next.js · Redux Toolkit · Ant Design · Shadcn · Ruby on Rails · PostgreSQL · Redis · Node.js · GraphQL · Elasticsearch · Docker · GitHub Actions · Vercel — mostly for North American startups.
VerifyState machines · covering arrays · model checking · property tests · database constraints
AgenticClaude Code · OpenAI Codex · Model Context Protocol · GitHub Spec Kit · Vercel AI SDK · Braintrust
NoteListed because they're real, not because they're the point. The point is upstairs.
Contact
Hiring for agent reliability, evals or full-stack work? Write to me.
Open to research and applied-AI engineering roles in the US, with visa sponsorship. Part-time contracts too.