Daily Tech News 20 February 2026

• The US government is building a specifically VPN-enabled website to allow Euroserfs to access content banned by their fascist overlords. (Reuter)
  
  Radio Free Europe, part two. Only this time the Europeans aren't being oppressed by the Soviet Union, but by themselves.
  
  
  
• Why Europe doesn't have a Tesla. (Works In Progress)
  
  The article highlights the continent's absurdly restrictive labour laws which are certainly a key part of the problem.
  
  But European leaders have decided on a slow and messy suicide, and many of their voters are entirely on board with this.

• PromptSpy is the first known Android malware to use generative AI at runtime. (Bleeping Computer)
  
  Which is interesting, because it uses Google Gemini and that's not free. With a bit of counter-hacking you could bankrupt the bastards.
  
  
  
• Very not free, in fact. (The Red Beard)
  
  Although Google Gemini has nominally lower prices and a larger context window than (for example) Anthropic's Claude, it works differently and will very happily burn through all your money to construct the perfect answer to a question you never intended to ask.
  
  
  
• It's an expanded D&D dice set: AMD's Ryzen 10000 series due later this year is expected to come in 6, 8, 10, 12, 16, 20, and 24 sides cores. (Tom's Hardware)
  
  There was no reason AMD couldn't have put out a 10-core CPU (from Zen 3 onwards, anyway), it just wasn't terribly practical.
  
  Now, though, 10 is the new 6 and 20 is the new 12 - a pair of disabled cores per chip.
  
  
  
• The human behind the rogue AI bot "Crabby Rathbun" that has been trying to push unwanted patches into random Python projects has come forward - maybe - and apologised - sort of. (The Shamblog)
  
  It's depressing if accurate, because it points to a future not only of unending slop, but of increasingly aggressive slop generators.
  
  
  
• At least we can hack them. (The Register)
  
  That scene in the movie Independence Day where the alien mothership is destroyed by an uploaded virus suddenly seems a lot more plausible if they were running AI on their computers:

  Prompting each of them to generate 16-character passwords featuring special characters, numbers, and letters in different cases, produced what appeared to be complex passphrases. When submitted to various online password strength checkers, they returned strong results. Some said they would take centuries for standard PCs to crack.

  Sounds good.
  
  Too good to be true:

  The researchers took to Claude, running the Opus 4.6 model, and prompted it 50 times, each in separate conversations and windows, to generate a password. Of the 50 returned, only 30 were unique (20 duplicates, 18 of which were the exact same string), and the vast majority started and ended with the same characters.

  If your "random" password generator returns the same result 36% of the time, you have a big problem. And it's actually worse than that:

  The team used two methods of estimating entropy, character statistics and log probabilities. They found that 16-character entropies of LLM-generated passwords were around 27 bits and 20 bits respectively.

  20 bits of entropy is four lower-case letters. You don't need a supercomputer to crack that; with enough patience you could do it by hand.
  
  
  
• The most recent piece of technology I own is a 3D printer from 2024 and I keep a loaded gun ready to shoot it if it ever tries to report me to the government. (Adafruit)
  
  New legislation introduced in California would not only require 3D printers to match all designs against a central database before being permitted to begin work, but would require the printers themselves to be so registered, and would make it a crime to sell or transfer unregistered printers.

posted by Pixy Misa at 04:00 AM

| Access Comments