The Caterpillar Loop

Where feedback usually dies

On most teams, a bug spotted by a junior tester takes a long, lossy road to a fix: a note in a chat, a verbal hand-off, a triage meeting, a ticket written days later by someone who wasn't there. Detail leaks at every step. The person who saw the problem ends up three hands removed from the person who fixes it — and the fix is only as good as what survived the retelling.

The Caterpillar Loop deletes those hand-offs. The observation and the specification become the same act.

Two ways in, one inbox

Every report — whoever files it — lands as a labeled GitHub issue in one private repo. There are two front doors:

What the tester never has to think about — captured on every in-game report:

• screen
• question ID
• pack
• answer chosen
• live / offline
• client
• browser
• timestamp

The hand-off to Copilot

This is the hinge. Because each report arrives as a clean, fully-contextual GitHub issue, it's already in the exact shape GitHub Copilot's coding agent needs. Assign the issue to Copilot and it reads the context, works the code, and opens a pull request — a first draft of the fix, ready for review.

A junior QA and an AI junior developer, paired: she supplies the ground-truth observation, it supplies the first draft of the fix, and a senior review is all that stands between them and shipping.

Claudia, amplified

Claudia is our junior QA. In a traditional shop, a junior tester's ceiling is "finds bugs and reports them" — and the value of that report decays with every hand-off it survives. Here, her report is the first domino in an automated chain that ends in a code change. She isn't handing work to the pipeline; she's the one who starts it.

Every tap is a fully-specified issue an AI agent can act on immediately — zero translation loss between what she saw and what gets fixed. That converts her hours into shipped improvements at a rate a junior role almost never reaches. She costs a fraction of a senior engineer and, inside this loop, lands a wildly outsized share of the impact.

It reframes what junior QA even is. She's not the bottom of a ticket queue. She's the sensor at the front of an automated delivery line — the human judgment that decides what's worth fixing, wired straight into the machinery that fixes it.