The full teardown. Everything in the critical review, plus the questions you can only answer by reading the whole thing: does this architecture survive real users, is the data model going to trap you in six months, and is it cheaper to fix this or rebuild it?

You'll get an honest verdict on that last one. Sometimes the answer is rebuild, and you deserve to hear it from someone who isn't quoting you for the rebuild.

Architecture review: where the AI made structural choices that don't scale, and what they'll cost you later

Data model and schema review — the migrations you'll regret

Load reality check: what breaks at 100 users, 1,000, and 10,000

Code quality: duplication, dead code, hallucinated or unmaintained dependencies, inconsistent patterns across AI sessions

Dependency and supply-chain audit

Test coverage assessment and a ranked list of what to test first

AI/API cost modelling at realistic usage — the bill nobody forecasts

Mobile responsiveness and baseline accessibility check

Prioritized remediation roadmap with effort estimates per item, so you can hand it to any developer

Repair vs. rebuild recommendation with reasoning

Kickoff call and findings walkthrough

Two weeks of async follow-up questions after delivery

Scope: One codebase, up to ~50,000 lines. Findings, roadmap, and estimates, implementation quoted separately.