@leo_linsky@PussyBlaster10k@LH There is no audience for your garbage benchmark. Everyone building real things knows the harness and prompts need to be tuned to the model. There’s no value in a benchmark that holds the harness and prompts constant across between models.
@kushbhuwalka They also have a pretty harmful mindset of believing they’re still innovative when they’ve been months behind for months.
All their tweets are acting like they just discovered sliced bread, when it’s something everyone else has already been doing.
@MarcBeirne@tenobrus@Dragonogon > probably not related to whether you're still using SO
The 2023 survey had 90k participants, the 2026 survey had 30k participants. They lost 2/3 of their participants in the last three years. It’s obviously correlated with Stack Overflow use.
@MarcBeirne@tenobrus@Dragonogon 1. Not sure what you’re trying to say by linking to that.
2. The Stack Overflow survey is conducted through SO’s channels, which makes it pretty useless in 2026 and especially useless for anything involving AI, since the only ones who still use SO are a tiny group of luddites.
@consentengineer 1. local elections are nearly always on the same ballot as national elections, so you’re not saving any effort by abstaining from national while participating in local
2. the more smaller the election, the greater the share
3. yet here you are spreading your belief 😉
@brandur This is the original sin of https://t.co/HPu4T26C7l. They did not understand this and it took them forever to realize their mistake.
Same with all the other “deploy your servers where your users are” hosts.
@louisagareiss@rizzu26 Fair, though I suspect the U.S. market is better for finding software engineering work.
I don’t know what the EU market is like, but in the U.S. needing a visa for an entry/intern level role makes it extremely difficult.