I have simply cared about "optimizing output" long ago. Ironically, I think I ship a lot more stuff than people who waste all their time optimizing their life away.
Somehow in the Netherlands it seems we have mostly managed to escape this. But it's just because people didn't really put any decoration on their menus before, and no decoration is still cheaper than shit AI decorations.
I got one of those random 2auth codes email and I assumed my password had been compromised. At least it's some kind of relief to know that it's only a compromised Microsoft email address...
On my side, I can give many examples of random software that became significantly worse since the AI trend started.
Trainline is practically unusable for purchasing itineraries that go accross multiple european countries. GitHub Actions now contains a bunch of extremely frustrating random bugs. Grammarly somehow gives worse copy recommendations.
It's a watch. It's an item that literally has one single job: tell the time accurately. It goes without saying that it should do a pretty good job at this and be accurate.
When I was a teenager, tiny core saved me for a few months. My laptop had died and all I could use until I got a replacement was an old desktop computer we had around with 256MB of RAM. It was around the end of the windows 7 era, so even Xubuntu was struggling on such an old computer.
Tiny Core ran surprisingly well and I could actually use it to browse the web and use IRC.
Not only just one run per model, but no metrics other than total return. If you pick stocks at random you have a very high chance of beating the S&P 500, so you need a bit more than that to make a good benchmark.
This resonates with me. Last week I got stuck on a bug where GitHub actions was pulling ARMv7 docker images when I specifically requested ARMv8. Absolutely impossible to reproduce locally either.