Architecture triage
I separate useful structure from AI cosplay before the app ossifies around bad decisions.
You can read the code. You still need a closer.
You are technical enough to know the AI output is not all wrong. That is the problem. The broken parts hide inside plausible patterns, fake architecture, and tests that prove nothing useful.
Send the repo-shaped problem
A sharp brief is enough to start.
Where I help
Debugging loops
I track the actual boundary instead of prompting around the same symptom for six hours.
Production hardening
I tighten routes, state, data flow, error handling, and deployment checks until the app behaves.
The engagement
- 01
Fast repo read with the failure path front and center.
- 02
A decision list: keep, cut, fix now, fix later.
- 03
Focused implementation or pair-style review on the part that keeps slipping.
Good fit
- You know enough to be dangerous and want sharper judgment.
- You need someone who can work inside the code, not around it.
- You care about shipping without pretending the mess is fine.
Bad fit
- You want validation for every generated abstraction.
- You refuse to delete code the model sounded confident about.
- You need a cheerleader more than an engineer.
Use AI for leverage. Use me for judgment.
Send the failure path and the repo. I will help you stop negotiating with plausible nonsense.