Run an AI Code Review Before the Human One

Background: Native [iOS / Android] app, [architecture], [minimum OS].
Team conventions worth knowing: [threading rules, error handling,
whatever your team cares about]. Here is the diff:

[Paste the changed files, or the PR diff]

Use case: Review this the way a demanding senior mobile engineer would,
before I put it in front of one.

Implementation details:
- Correctness first: threading and main-thread violations, lifecycle and
  scope handling, retain cycles or leaked coroutines, force-unwraps and
  null handling, state that can get out of sync
- Then mobile-specific concerns: does this work offline, on a slow
  connection, on a small screen, with the largest accessibility font, in
  the background, on a cold start
- Then fit: does it match the architecture and conventions above
- Then tests: is what's tested the part that can actually break
- Rank findings by severity and say plainly which ones you'd block on

Deliverable: A prioritised review with file and line references,
severity, the concrete failure each finding causes, and a suggested fix.
Separate "this is a bug" from "this is taste"—and say which of your
findings you're least sure about.