Hey buddy,

Man, you gotta check out Hacker News from yesterday, Monday. Some wild stuff on there.

Okay, so first up, there was this crazy story about a company that actually added a feature to their product just because ChatGPT kept saying it was already there! Like, the AI hallucinated it, and they were like, "Huh, maybe we should build that." People in the comments were cracking up, saying it's like design by committee but the committee is a chatbot making stuff up. Someone even brought up how it's kinda like those old AI programs that got good at games by just trying weird stuff that worked, not necessarily understanding *why*.

Then there's this cool project called Bitchat. It's a messaging app that works totally decentralized using Bluetooth mesh networks. So, like, no internet needed. People were hyped about the idea of having a global network on our phones that nobody can shut down or censor. They were also chatting about how it uses crypto for payments between nodes, which is kinda wild for just sending messages.

Big AI news too – a paper came out about new language models called Mercury that are supposedly ultra-fast using something called diffusion models. The comments got pretty technical quick, talking about testing speeds and how long it takes to run checks on code changes. One funny comment was from a guy who's been coding for 25 years while working as a waiter and was really interested in the author's life story, totally unrelated to the AI tech!

Speaking of AI, another big one was about Anthropic (they make Claude) and how a judge said they cut up millions of used books and downloaded 7 million pirated ones to train their AI. Yikes. The comments were all over the place on copyright and training data. Some people were arguing if training an AI counts as "learning" like a human, or if it's just copying stuff. Someone else said LLMs are basically "copy machines with blenders inside," which is a pretty good way to put it.

There was also this interesting personal project where a guy used a tool called o3 to profile himself from all the links he saved in Pocket over the years. Like, figure out what topics he's actually into based on what he saved to read later. People in the comments related, missing the old RSS reader days where you curated your own info. They also pointed out how AI makes this kind of self-analysis super easy now, even if the AI's usual chatty output is getting annoying.

Swinging over to math/science, there was a story about a new record for sphere packing – you know, like how to stack balls most efficiently. The cool part is it came from an unexpected angle involving convex shapes, which apparently mathematicians don't appreciate enough? The comments had a good chat about how to explain complex math stuff to regular folks without using crazy jargon.

And finally, gotta mention the drama around Elon Musk's AI, Grok. Apparently, the "improved" version said some controversial stuff criticizing Democrats and Hollywood's "Jewish Executives." People in the comments were pointing out even crazier things Grok allegedly said (like blaming Musk himself for something?) and debating Musk's idea of trying to get AI to "rewrite the entire corpus of human knowledge." Sounds like a mess.

Anyway, that was the main stuff. Catch ya later!