The One Red Run

I keep seeing posts like this one.

Wonder if your tests are helping? Ask Claude.

Prompt: “Look through my past sessions. How many times did a test catch a bug?”

I just asked Claude and it found 5 recent instances where a test caught a bug.

Seen on X, October 2026

My first thought was... when have you ever done this before AI?

describe('counting how often your tests caught a bug')

it('was never a question before AI')

I can't think of anyone who ever sat down and counted how many times their tests caught something, just to see if they were worth having. We could have. Nobody bothered, because the number doesn't really tell you anything.

it('only needs to go red once')

Tests aren't written to catch things all the time. They're there for the one time something does break. Same as a seatbelt really. You don't stop wearing it because it didn't do anything last year.

1,055 boring runs
and one that mattered.

A test that's never gone red might be useless. Or it might be guarding something nobody's touched yet. The count can't tell the difference, and that's exactly when you'd be using it... to decide what to delete.

it('points at the wrong tests')

If anything it points the wrong way. The tests that go red the most are usually the brittle ones, the ones that break every time you refactor even though nothing's actually wrong.

it('says more about the agent than the tests')

And the five? That's mostly the agent breaking something and a test catching it before it went anywhere. Which is great, and honestly the bit worth talking about. Agents change a lot of code quickly and start every session with no idea why things are the way they are (more on that in Every Session Is Day One), so tests matter more now than they did.

But five catches tells you how often the agent broke things. Not how good your tests are.

it('asks a better question')

If you want Claude to help, get it to break things on purpose and see if anything goes red. If nothing does, you know what test to write.

AI's made it really cheap to ask questions we never bothered asking before. Sometimes we never bothered for a reason.