Hey buddy,

Just wrapped up looking at Hacker News for Wednesday, June 3rd, and man, there was some interesting stuff. Thought I'd give you the quick rundown.

AI and Tech Updates

First off, Google dropped their new AI model, Gemma 4 12B. Sounds like it's a big deal, a "unified, encoder-free multimodal model" or whatever. People in the comments were saying the 31B version is super impressive and you can get a bunch of free requests, but also that it gives a lot of errors. Someone also mentioned that for laptops with 16GB, other models like Qwen 3.5 9B are still the champions. So, new AI, but maybe not for everyone's home setup yet.

Then there's Elixir, the programming language. They just released v1.20, and it's now a gradually typed language. This is a pretty big deal for them. The creator jumped in the comments to explain why type systems are good, even in the age of AI coding, saying they help restrict programs and increase expressiveness. Someone else was really hyped about it being the first notable language to support "full set theoretic types" – unions, intersections, and even complements. Sounds pretty advanced!

Privacy and Hacking

This one's wild: Meta workers can apparently opt out of being tracked at work for up to 30 minutes. Only 30 minutes, though! The comments were discussing how tricky workplace tracking is, especially with laws in different countries. A common piece of advice was, "don’t work on personal projects on work machines," because your company might claim ownership. Good reminder!

And check this out: a post called Pwnd Blaster: Hacking your PC using your speaker without ever touching it. Turns out, some guy figured out how to reflash a Creative Sound Blaster Katana V2X soundbar's firmware via Bluetooth without any real authentication. So, not just *any* speaker, but still a pretty scary vulnerability for that specific device. People in the comments were saying it's like leaving an S3 bucket open, but for Bluetooth devices. Pretty clever, but also terrifying.

Oh, and there was a pretty serious personal story about someone who was diagnosed with anti-NMDA receptor encephalitis. It's a rough read, but important. The comments section was full of people sharing similar experiences about being misdiagnosed for years, especially women, with some mentioning "medical misogyny" in how symptoms are often dismissed. Really highlights how hard it can be to get proper medical care for complex conditions.

AI's Impact and Costs

There was a cool discussion around Uber's $1,500/month AI limit, which people are calling a useful signal for how AI tools might be priced in the future. Someone pointed out that current API prices from big players like Anthropic and OpenAI are probably massively inflated. It's making people think about whether it's cheaper to run their own LLMs eventually, or if we'll see more regional AI providers pop up.

Finally, a thought-provoking piece called "They're made out of weights" was talking about how AI models work, using a fun analogy. It's basically saying that the 'magic' of AI isn't some mystical thing, but just a huge collection of numbers (weights) that represent learned patterns. The comments dove deep into consciousness and what it means for an AI to "understand" something. Definitely some food for thought there.

Anyway, that's the gist of it. Hope you're having a good one!