← Voices / AI in action QA Owning the loop

Twenty-Three Hours

A server error only some countries could trigger — and the day we stopped throwing bugs over the wall.

by The Wanderer · June 2026 · 4 min read

Inspired by real events. Part real, part retelling — the way you'd tell a story to a friend. Names and details may have been changed or interpreted. If this caused any discomfort, we're sorry.

Yesterday a bug came in. Nothing about it looked special — a user report, the kind that lands in the queue a dozen times a week.

Base Itinerary Trip Search was throwing a Server Error. Not everywhere. Not for everyone. Just for some countries — Australia, Thailand — and only on Artemis. Classic hard-to-spot edge case. The kind that, in the old world, meant a week of back-and-forth before anyone even understood why.

Here's what made this one different.

The Closer didn't just log the ticket.

For as long as I've been around software teams, the shape of this was fixed. QA finds the bug. QA writes it up. QA throws it over the wall to engineering. And then everyone waits — for someone with enough context on that corner of the system to pick it up, page in the history, find the cause, and write the fix.

The Closer skipped the wall.

He sat with Claude and walked the analysis until he understood the root cause himself — not "here's where it errors," but "here's why it errors, and why only for these countries." And then he wrote the fix.

The win wasn't the speed. It was that the person who found the problem understood it well enough to solve it.

The timeline tells the rest of it on its own:

the real shift QA's job was never finding bugs. Finding was always the easy part. The wall was the diagnosis — the wait for someone who held that part of the system in their head. When the person closest to the problem can also reach the cause, the wall just… isn't there anymore.

Now the honest part, because this site runs on honesty.

This was one clean case. Not every bug closes in 23 hours. Some edge cases are genuinely tangled. Some fixes ripple into three other things. Some root causes hide for days. I'm not going to stand here and tell you we've abolished the hard bug — we haven't.

But this wasn't a fluke either. It was one person owning a problem from the first symptom to the last deploy, with a tool that let him hold the whole system in his head at once. That part is repeatable. That's the part worth writing down.

what kept it moving None of this works solo. The Gatekeeper and The Co-Signer reviewed and approved fast enough that the momentum never broke. A 23-hour fix needs a 23-hour review. The loop is only ever as quick as its slowest gate.

Big shoutout to The Closer for owning it. And to The Gatekeeper and The Co-Signer for the reviews and approvals that kept it moving. 🙌

The bug was flagged on a Tuesday. By the time I read the thread the next afternoon, it was already closed. Already shipped. Already quietly serving travellers in Australia who'll never know it broke.

No war room. No week of waiting.

Just a QA engineer who decided the wall wasn't his to respect.

⏭
If this one landed, read the why behind it: When AI collapses the cost of writing the code, the bottleneck moves from technical execution to understanding the problem — and a new kind of contributor steps in to fill the gap.
The Business Technologist →
about the author
The Wanderer was one of the first in the cohort to embrace vibe coding and now lives and breathes this stuff as an AI PM. He likes the stories where the workflow quietly changes shape and nobody makes a fuss about it.

Have your own AI story? WhatsApp Sree →  ·  or email →