Man, you wouldn't believe the stuff on Hacker News today. It was a pretty wild mix, especially with all the AI chatter, as usual.

Crazy Tech Business Card

First off, get this: someone made an ultrathin business card that actually runs a fluid simulation. Like, a real-time fluid sim on a card! Super cool project, but the comments were hilarious. Everyone was dogging on the font on the back, saying it looked like an unstyled HTML page from a research professor. Apparently, white text on black is harder to read, so the font needed to be bolder. And some folks were talking about how they get business cards now – less about work, more about shared hobbies. Wild, right? Here's the link if you wanna see it: https://github.com/Nicholas-L-Johnson/flip-card

Building an Offline AI Lab

Then there was this dude who wrote about wanting everything local and building his own offline AI workspace. You know, no cloud, all on his own machines. He's trying to get away from relying on big tech. The comments had a big debate about Macs and running LLMs – some people swear they're good enough, others are like, "nah, you're just impressed it runs at all." It's all about that local-first dream, I guess. Check it: https://instavm.io/blog/building-my-offline-ai-workspace

Farewell to an Apollo Legend

On a more somber note, Jim Lovell, the commander of Apollo 13, passed away. That's a huge loss. You know, the guy who said "Houston, we've had a problem." The comments were mostly respectful, but one cool detail popped up: apparently, David Scott, the commander of Apollo 15, was the technical consultant for the Apollo 13 movie. Ron Howard wanted it super accurate, down to the smallest detail. Pretty neat. Link here: https://www.nasa.gov/news-release/acting-nasa-administrator-reflects-on-legacy-of-astronaut-jim-lovell/

Why Can't *I* Run GPT-4 Locally?

Back to AI: there was this popular "Ask HN" post, "How can ChatGPT serve 700M users when I can't run one GPT-4 locally?" It's a question we all ask, right? The answers were super insightful. It's all about batching requests, making inference stateless, and routing small amounts of data to massive machines. Basically, they're not running one giant model for each person; they're super efficient with how they handle requests. Makes sense, but still feels wild. Here's the thread: https://news.ycombinator.com/item?id=44840728

GPT-4o Drama

And speaking of ChatGPT, there was a whole thing about the "surprise deprecation of GPT-4o for ChatGPT consumers." People were PISSED. Apparently, OpenAI just swapped it out or changed access without much warning. Some of the comments