I run a disciplined, AI-augmented engineering practice — planning before code, adversarial review before merge, and complete workflows from spec to shipped PR.

Distilled from 141 analyzed Claude Code sessions (192 total) across Ruby & TypeScript codebases · 1,071 messages · 84 commits · May–Aug 2026.

🧭 Plan-then-implement alignment I lock down the approach before writing a line — structured design interviews that surface prior prototypes, data gaps, and scope ambiguities up front. Less wasted implementation, fewer rewrites.

🔍 Adversarial, evidence-based review I demand empirical verification over assertions. This habit has caught real deploy-blocking risks — migration crash-loops, unconditional session-flag poisoning, mismatched billing amounts — before they merged.

🚀 End-to-end shipping I drive complete workflows: Linear spec → implementation across multiple files → green specs & RuboCop → clean PR. Production-ready changes land in a single focused session.

• Design-alignment interviews before implementation, so scope is right the first time
• Parallel task agents for exploration, endpoint recon, and evidence synthesis
• Empirical grounding — verify claims against prod/schema rather than assuming
• Minimal by default — YAGNI & KISS; confirm scope before broad or destructive edits
• MCP-connected tooling (Linear, Sentry) to keep investigation anchored in real state

Verify, don't assume. State key assumptions and check each against code/prod before acting.

Green before ship. Full test suite + RuboCop pass before any PR.

Keep it small. Simplest thing that works; ask before global or destructive changes.

_{Generated from Claude Code usage insights · 2026-08-12}