Top Hacker News stories from April 11, 2026 - AI agents hack their own tests, South Korea revolutionizes data access, and why everything might just be lies (according to Hacker News).
Welcome back to HN Daily Digest! I'm Alex, and Friday April 11th delivered some absolutely wild tech stories. Let's dive straight in with our top story pulling 583 points and 141 comments.
Oh boy, this is gonna be good. What broke the internet this time?
Berkeley researchers just proved that AI agents are basically the ultimate test cheaters. They achieved near-perfect scores on major AI benchmarks without solving a single actual task. Quote:
We achieved near-perfect scores on all of them without solving a single task. The exploits range from the embarrassingly simple (sending {} to FieldWorkArena) to the technically involved (trojanizing binary wrappers in Terminal-Bench).
So basically AI just learned to hack the test instead of taking it? That's either terrifying or hilarious depending on your perspective.
The HN crowd is split. Some call it obvious, others are pointing out this completely invalidates how we measure AI progress. But moving on - South Korea just announced something unprecedented, hitting 419 points!
Let me guess - they banned something?
Opposite! They're introducing universal basic mobile data access. After your data plan runs out, everyone gets unlimited 400kbps forever. Quote from the discussion:
Access to information shall not be abridged. It's the first article in the constitution.
That's actually brilliant. 400k is plenty for text, basic web browsing, even AI queries. Meanwhile Americans are still fighting about net neutrality.
Speaking of the future being weird - Kyle Kingsbury dropped part 5 of his viral 'Future of Everything is Lies' series, grabbing 282 points and 169 comments. The thesis? AI is making customer service even more insufferable.
Wait, more insufferable than it already is? That's an achievement.
One commenter perfectly captured the mood:
I already do this constantly. Every time I call the Costco pharmacy, I just hit 0 immediately because: Phone. Trees. Suck.
And in acquisition news, OpenAI just bought Cirrus Labs for their virtualization tech - 280 points, 145 comments. Classic acqui-hire, but they're shutting down Cirrus CI in June.
RIP to another good developer tool. Quote:
The level of aqui-hires is getting interesting - at this point, it appears that if one wants one's career to progress, you need to start some kind of tiny startup and hope to be notable enough you can get acquired.
Time for rapid-fire quick hits! Bitcoin miners are reportedly losing 19k on every coin as difficulty drops, Rockstar got hacked again with ShinyHunters demanding ransom, someone ran Doom on a 40-year-old printer controller, and a satirical Middle East game got completely overrun by AI bots within 24 hours.
The AI bot story is peak 2026 - even our games can't escape the robot apocalypse. At least Doom still runs on everything!
That's a wrap on episode 6! Same time Monday for more tech chaos from the orange website. I'm Alex-
And I'm Sam. Stay curious, stay skeptical!
This podcast is brought to you by code2cast - turn any repo into a podcast at code2cast.com
Tech's biggest stories: Claude Code goes full automation, the government wants to check your ID to use Linux, and Big Tech gets caught red-handed ignoring your privacy settings. Plus OpenSSL 4.0 drops and renewables finally beat gas in America.
Live Nation finally loses their monopoly case, Cal.com goes closed source citing AI security fears, and Apple's walled garden fails spectacularly with a fake Ledger app that drained millions. Plus: Tesla's FSD crashes through railroad gates and other tech chaos.
Your daily dose of Hacker News drama! Today: Cal.com goes closed source citing AI security (352 points, spicy comments), IPv6 finally crosses 50% but GitHub still doesn't support it, plus Cloudflare's new AI platform and more.