Every fleet app reports green about itself. This one scans them all from outside with the same yardstick, so lost coverage shows up as a regression, not a guess.
Each one answers a different part of "is this code actually proven?" — across the fleet, every night.
Nightly, each app's default branch runs in a throwaway worktree with a probe that records which routes, Livewire actions and methods every test reaches.
Every surface is marked Tested, Weak or Missing. Tested methods get their body mutated — a surviving mutant downgrades the verdict to Weak.
Each scan diffs against the last. Lost coverage and new untested surfaces are regressions — named, not averaged away.
An AI agent writes tests for gaps, verifies them itself — reach, mutation, full suite — then opens one PR per app for a person to review.
Every test of every app — read straight from git, with its revision history, outcome over 30 scans and how many surfaces it proves.
Scripted personas walk the real UI in a headless browser. Failed steps, server errors and clicks with no loading feedback become findings.
Every stage of a customer's life — from first enquiry to cancellation — walked across every app it touches in a stand-in fleet.
Built
Built
Built
Built
Built
Pending
Built
Built