Just wanted to give you a quick heads-up on some interesting stuff that popped up on Hacker News today, Tuesday, July 29, 2025. Lotta AI talk, as usual, but some other cool bits too.

OpenAI's "Study Mode" for ChatGPT

First off, OpenAI is launching something called "Study Mode" for ChatGPT". Sounds like they're trying to make it a serious learning tool, not just for quick answers. Like, imagine ChatGPT being your personal tutor for school or learning a new skill. The comments were pretty mixed, with some folks saying Google already has a similar thing called "leanlm". Someone also brought up a good point about what a "proper" learning tool needs, like tracking your progress and understanding your knowledge level, instead of just spitting out info.

Laptop Writing Space Invaders with Local AI

This one's super cool: a guy got his 2.5-year-old laptop to write Space Invaders in JavaScript using a local AI model (GLM-4.5 Air). Seriously, a whole game! It shows how powerful these AI models are getting and that you don't always need a super-computer or cloud service to run them. One of the comments asked if this means proprietary AI models (like the big ones) are becoming less relevant now that these open-source, local ones are so good. Food for thought, right?

Cats Confusing AI

Speaking of AI, there was a hilarious one about how adding "irrelevant facts about cats" to math problems increased LLM errors by 300%. So, if you say "Johnny has 10 apples, and a fluffy cat sat on 5 of them, how many does he have?", the AI just completely melts down. It really highlights how these models aren't "smart" like humans; they just get thrown off by extra noise. Someone in the comments even pointed out that some kids make similar mistakes, which is kinda funny and sad at the same time.

Wikimedia vs. UK's Online Safety Act

On a more serious note, the Wikimedia Foundation is challenging the UK's Online Safety Act regulations. This is a pretty big deal for internet freedom and privacy. Basically, Wikipedia is saying "hold on a minute" to the government wanting more control over online content. A lot of people in the comments were slamming the act as massive government overreach, saying it's a "sad attitude for a forum called HackerNews" if anyone defends it.

"Unlimited" Doesn't Mean Unlimited

Then there's the classic tech gripe: "Stop selling 'unlimited', when you mean 'until we change our minds'". This article was about companies offering "unlimited" services (like AI usage or storage) and then quietly putting limits on them later. Super annoying bait-and-switch, happens all the time. One of the comments mentioned that Australia actually outlawed advertising "unlimited" broadband because it was so deceptive. Wish more places would do that!

EU Could Be Scanning Your Chats

Finally, a pretty concerning one: "The EU could be scanning your chats by October 2025". This is about potential new regulations that would allow authorities to scan private messages for illegal content. Obviously, a huge privacy concern for everyone. People in the comments were talking about how some child safety scanning already exists, but expanding it to general chat is a whole different ballgame. One user even suggested using something like a private SSH server for secure chat, which is a bit extreme but shows the level of worry.

Anyway, that's the quick rundown. Talk later!