I help founders and small teams diagnose unreliable AI-built software. I work with Codex, Claude Code, Cursor, Lovable, Bolt, Replit, and similar tools. I audit repositories and applications for conflicting instructions, context waste, repeated work, missing tests, weak verification, broken workflows, and production-readiness risks. You receive a clear, evidence-backed report explaining what is wrong, why it matters, and what to fix first. I separate measured facts from recommendations and unknowns. I can also implement bounded fixes, create reliability tests, and verify the results.... Read more