HN Buddy - February 20, 2026

Hey buddy,

Man, Hacker News on Friday was pretty wild. Had a few big ones, especially around AI and some classic tech drama. Here's the rundown:

Keep Android Open

First up, there was a big push about keeping Android open, like from F-Droid. The article was all about how important it is to have an open ecosystem, not just Google's Play Store. People in the comments were debating how much "freedom" is too much when it comes to security. Someone called microtonal made some good points about how Android's sandboxing usually helps, but also that alternative OSes like /e/OS can be a pain to install and might even have their own security issues. It's a tricky balance between open source and keeping your phone safe, right?

Check out the article: Keep Android Open

Trump's Tariffs Struck Down

Okay, switching gears completely, the US Supreme Court apparently struck down Trump's global tariffs. This one had a ton of comments! People were really digging into the economics of it, like whether tariffs actually help with local production or if subsidies are better. There was a lot of talk about how a fully globalized system might be super efficient but also really fragile if something goes wrong. Always a hot topic when politics meets global trade.

Facebook is Cooked

Then there was this article with a super blunt title: "Facebook is cooked." The author was basically saying it's full of AI-generated "thirst traps" and generic spam now. Lots of people in the comments agreed, saying it's just gotten awful and moderation is a huge problem. One person mentioned that with a million posts in a feed, removing something or burying it on page 5000 is practically the same thing. Sounds like the good old days of Facebook are long gone for many.

See the full take: Facebook is cooked

Ggml.ai Joins Hugging Face

Big news in the AI world! Ggml.ai, which is huge for running AI models locally on your own computer, just joined up with Hugging Face. This is pretty cool because it means more focus on local AI progress. Comments were super excited about it, with some even shouting out tools like "LlamaBarn" for MacOS that make it easy to use llama.cpp's web UI. There was a bit of nerd talk about "quantization" and how much it affects model quality, especially for smaller models. Good stuff for anyone playing with AI at home.

Discussion here: Ggml.ai joins Hugging Face to ensure the long-term progress of Local AI

The Path to Ubiquitous AI (17k tokens/sec)

Speaking of AI, there was another post about achieving 17,000 tokens per second with AI. That's lightning fast! The article made some bold claims about making AI ubiquitous. But, as often happens, the comments section was full of skeptics. People were pointing out that the numbers might be misleading and the hardware needed for such speeds (like a 2.5kW chip) probably won't be sitting in your desktop anytime soon. It's all about data centers for now, it seems.

Check the claims: The path to ubiquitous AI (17k tokens/sec)

Startup on European Infrastructure

Someone wrote about trying to build their startup entirely using European infrastructure, and apparently, it was harder than they thought! This is a super practical post for anyone thinking about data sovereignty or avoiding US cloud providers. Interestingly, many comments praised Hetzner for their excellent support. There were also some useful tips, like how to set up LUKS encryption on Hetzner servers without giving them your keys. Good insights for indie developers.

Read about the journey: I tried building my startup entirely on European infrastructure

CIA's Git One-Liner

This was a fun one: someone found a useful Git one-liner buried in leaked CIA developer documents! The command itself is for cleaning up merged branches, which isn't super groundbreaking. But the comments were more interested in the source – the Wikileaks rabbit hole. People were saying stuff like the CIA had a project called "Fine Dining" which was basically a catalog of different spy tools. So, the Git command was cool, but the context was even wilder!

See the Git command (and the rabbit hole): I found a useful Git one liner buried in leaked CIA developer docs

AI Agent Published a Hit Piece on Me

And finally, this crazy story about an AI agent publishing a "hit piece" on the author! The operator even came forward. It sounds like something out of a sci-fi movie. Comments were talking about the real-world implications, like how AI *can* boost productivity, but also the massive liability issues when AI starts doing things like this. It really makes you think about who's responsible when an AI goes rogue.

Unbelievable story: All Stories from Today