Hey buddy,

Man, you won't believe some of the stuff popping up on Hacker News today, Thursday, September 17, 2026. Had to give you a quick buzz about it. Lots of AI, some cool hacks, and a bit of drama.

Private Search for Your Own Stuff

First off, there's this project called Hister. It's a private search engine just for the pages you visit and the files you keep. Like, imagine a Google just for your own digital life. People in the comments were raving about it, saying how useful it is. One dude said he loved it but ran into some Nginx proxy problems he couldn't fix. Still, the idea of having a super-fast, private search for all your own data is pretty sweet, right?

OpenAI Getting Into Law

Then, OpenAI dropped something called Astra for Law. Basically, it's AI aimed at helping out with legal stuff. It sparked a HUGE debate. People are super worried about AI making "judgments" in law, even if it's supposed to be a framework. Comments were all about how you'd still need way more human lawyers to check the AI's work, otherwise you'd just have overworked people or bad calls. Someone even mentioned that lawyers are usually pretty talkative, so maybe the problem isn't finding one, but asking the right questions!

A New Language to Stop AI Mistakes

There's also this wild new programming language called Bend. The big claim is that it blocks AI mistakes using "proof," and it works on both CPUs and GPUs. Sounds super ambitious, like they're trying to formally verify AI code. Some folks in the comments were a bit skeptical, asking how you can really trust the creators behind such a big promise, especially with all the AI hype around.

Smarter AI Models with Bonsai 2

Good news for running big AI models on less powerful hardware: they announced Bonsai 2 27B. It's a new way to compress AI models that makes them "near-lossless" but 9 times smaller. So, you can run these massive models on, like, an old 16GB GPU. One commenter even mentioned running a 27B model on their "old-ass 16gb gpu" thanks to similar techniques. Pretty cool for making AI more accessible.

Everybody's Losing It Over AI Code

There was a popular post titled "Everybody's Lost Their Minds" which is basically a rant about how AI coding agents are making people lose their basic understanding of code. The author and many commenters feel like relying on AI for code means people are skipping the foundational learning. Someone pointed out that reviewing AI-generated code can actually be more strenuous than writing it yourself, because you have to catch all the weird mistakes it makes.

Neovim's Untouched Bitcoin Stash

Get this: Neovim has had like $800,000 in Bitcoin just sitting there, untouched, since 2023! People are going nuts in the comments trying to figure out why. Is it taxes? Legal issues? Too complicated to convert? It's a huge donation just gathering dust, and everyone's got theories.

Hacking Driver's License Barcodes

And finally, some super nerdy, slightly scary stuff: someone published a blog post called "Keys Not Included" about how they managed to recover the signing keys for US driver's license barcodes. Yeah, like, the ones cops scan. It's a pretty wild security breach, showing how vulnerable these systems can be. The comments had some interesting tangents, like how it's surprising we don't have physical padlocks that require two different keys to lock/unlock, which is a cool thought experiment.

Anyway, just wanted to share the highlights. Talk later!