Thank you for the comment!
You can see the parts in the pictures in the post. Yes, as far as I can tell, only a few layers were printed, then I lost adhesion. The build did stay on the tank's floor, from which I had to scrape it off. The machine kept going until "finished", but did not print anything for more than 90% of the printing process.
I will try angling the parts next time, did not think it was necessary since the first test print, which was not angled, went so well.

Thanks for the suggestion! Just trying to understand, why would reducing the exposure time for the bottom layer help with adhesion?

It completed, but was stuck to the resin tank. No errors this time.

True, GPT does not return a "yes" or "no" 100% of the time in either case, but that's not the point. The point is that it's impossible to say if GPT has actually gotten better or worse at predicting prime numbers with their test set. Since the test set is composed of only prime numbers, we do not know if GPT is more likely to call a number "prime" when it actually is a prime number than when it isn't. All we know is that it was very likely to answer "yes" to the question "is this number prime?" in March, and very likely to answer "no" in July. We do not know if the number makes a difference.

I would steal this argument, but if it can be reposted here for free, then I don't think anybody really owns it. 🤔

I think it's called "Buckles" in English. https://en.wikipedia.org/wiki/Buckles_(comics)
"Chiffon" in French.