Hacker News Reader

Do they even have automated testing? This should have been detected not only in one layer of the release process, but in serveral ones.

View on HN

Not anymore. The test lab and testers got laid off. Now they test in production on end users.

So there’s an argument to be made that when the same person designs, implements and tests the change, any purported defense in depth is largely illusory.

Once upon a time I disagreed with this argument, insofar as I believed it wasn’t necessarily true as long as you had good peer review every step of the way.

But the shift to AI coding has made good peer review a heck of a lot harder to pull off.

It can create tests to make sure that things keep working.

For 100% of the root cause analyses for critical production defects that I’ve done over the past 18 months or so, a key finding was that AI-generated tests asserted that the defective behavior was correct.

Reply on news.ycombinator.com