Man, you won't believe some of the stuff from Hacker News today. Grab a coffee, this is quick.

Oxide and LLMs in Hiring

First up, Oxide Computer, you know, that cool hardware company? They posted about how they're using LLMs. What's wild is they're even using them a bit in their hiring process. The comments section went off, with some people who applied and got rejected saying they felt like there was a lack of human interaction. One guy even said he got an automated rejection after two months. Ouch. Oxide's CEO actually jumped in to say they do read applications, and LLMs aren't a substitute, but it still sparked a big debate about how fair this whole AI-in-hiring thing is.

Germany Goes Open Source

Then there's this news from Germany: the state of Schleswig-Holstein is ditching Microsoft completely for open source software. And get this, they're expecting to save millions! People in the comments were debating if Microsoft could actually "shut down" a nation's computers (which sounds a bit over the top, honestly) and how corporate environments aren't just "default Windows" anyway. Sounds like a big win for open source.

AI Still Hallucinating in Research

Remember how AIs just make stuff up sometimes? Well, it's still happening, and now it's in academic papers. There was a post about over fifty new "hallucinations" found in submissions for a big AI conference, ICLR 2026. We're talking fake citations and made-up people. Some folks in the comments were saying LLMs just aren't good enough for serious stuff if they can't even get facts right, while others were arguing about what "hallucination" even means in this context. It's a mess!

Claude Fails the Space Jam Challenge

Speaking of AIs, someone tried to get Claude (the LLM) to recreate the old 1996 Space Jam website. You know, the super basic, kinda ugly but iconic one? And it totally failed! The article was pretty funny about it. Comments were buzzing about whether LLMs are just "plagiarism machines" or if it's more about giving them the right, super-specific instructions. Someone pointed out that things like "responsive design" would trip it up, which wasn't even a thing back in '96.

Dollar Stores Ripping People Off

This next one is kinda infuriating: dollar stores are apparently systematically overcharging customers. Inspections found crazy high error rates, like 76% in some places! The really wild part, according to the comments, is that in places like North Carolina, the fines are so low ($5,000 per inspection) that it's actually cheaper for the stores to just pay the fine than fix their pricing systems. That's just wrong, man.

Garage-Built ICs!

Okay, this one's super cool for us nerds. Some mad genius actually fabricated an integrated circuit, the Z2, in a garage lab. Like, with lithography and everything! How insane is that? It's not a super powerful chip, obviously, but the fact someone did that without a multi-million dollar cleanroom is just mind-blowing. People in the comments were talking about how this is almost like an "inverse Moore's Law" for single-threaded performance and other projects trying to make chip fabrication more accessible. So cool.

Google's AI Long-Term Memory

And finally, Google is pushing forward with AI memory. They've got this new "Titans" architecture that's supposed to help AI have better long-term memory. The comments on this one were a bit of a tangent, though, mostly about IP theft and corporate espionage in the AI world, rather than the tech itself. Still, a big step for AI if they can get that memory sorted out.

Alright, that's the quick rundown for today. Talk soon!