Congrats on both this and your recent PB!

Good luck, just finished Chicago Marathon without training the last two months (due to injury) do not recommend. You got this!

Great, now it seems to work. I need to dig more into the specific violations, but so far so good.

Tried to run it, but got some errors trying to load the imap library. I assume GitHub issues is the right place for bug reports?

Poenget er ikke debatten, poenget er valgløftene. Synes det ville vært uredelig å gå ut med valgløfter om endring i skatt og så binde seg i et bredt forlik rett før valget som gjør at man ikke kan følge opp løftene sine.

Yeah, I agree it probably is a bit apples to oranges comparison, but found it interesting since 1.7x doesn't seem to absurd given the 2-3x on more "pure" tests, and the 1.26x total effect seems in line with the other samples as well. But real numbers from @push.cx would ofc be even better.

PS: This is given that the 1.7x increase for the non-io bound mentioned in your article is comparable to the 1.26x total increase.

Love this! But due to the increases in non-io speed from YJIT it actually seems some applications spend more than 80% waiting for IO. I did the maths for lobsters, and io bound seems to be more than 80% (up from ~65%). Calculations here: docs.google.com/spreadsheets...