Unanalyzed raw video. Internet noise. Those would qualify as, "a whole lot of garbage that someone could possibly, maybe analyze to make some statistical inferences about conceivably sensitive data."

There's not 10 petabytes of sensitive data in all the world. 10 petabytes is enough to store a copy of every movie and television show ever released to DVD plus every book ever written in any language on Earth.

What they captured was some sensitive data and a whole lot of garbage that someone could possibly, maybe analyze to make some statistical inferences about conceivably sensitive data.

They sabotage themselves just fine with user-hostile security implementations and bloat bloat bloat.

I'll bet. His case has less than no chance in court.

It's good. Until it isn't.

I'm a computer scientist with three decades of expertise in computer networking. There's this one AI-based job board that keeps trying to match me with delivery driver jobs.

Something like this needs density. If there's not enough people using it, then the per use cost will be far too much to make it economically viable. That makes cities much more attractive to startups like this. Of course there you have airspace issues with large buildings, so the true sweet spot may be relatively dense but very high income suburbs. But it sure as heck won't be rural.

Designing a structure with withstand a major act of war basically means digging it far enough underground that nothing can reach it. It was a stupid rule.

You have to check when they're available anyway. Even where people actually work "customary business hours," they take time off for things like doctors' appointments and childrens' special events.

I vote for option E: Everybody switches to UTC.

Local noon on the U.S. pacific coast will be 20:00 instead of 12:00 and the office day will start at 17:00 instead of 9:00 am. But in trade, I never have to do any mental gymnastics converting for someone else's time again.

You can do that with github and many of the others as well. My employer has a private github.

Except it really isn't. The stuff from 2 years ago is miles ahead from the stuff 5 years ago. The stuff from 1 years ago is yards ahead of 1 year ago. The stuff from 6 months ago is nearly indistinguishable from the stuff from a year ago. Diminishing returns is hitting hard and rapidly. Many experts think that LLMs are closing in on their technical limitations, regardless of how much extra compute its given. It will need entirely new techniques to actually become much more than it is now. Which may happen, but doesn't appear to be on the horizon.

A study you can't even directly link to? Yeah, I call bullshit.

And my personal experience is it's at least a 50% slow down. I have yet to ever have it do anything not completely trivial that wasn't badly insecure and broken, and when using them to provide static analysis the false positive rate is around 90%.

No, those who aren't willing are actually following the science. Every measurement so far, every actual study has shown AI code generation is 20% or more slower for senior engineers. Even scaleai, a company founded and run by Meta's AI chief, shows the same in their data (https://scale.com/leaderboard/rli). Possibly it will someday get there, but it sure as hell isn't ready yet.

Also, Android is an open source OS. You can always run it on a modified OS that pretends to have secure storage but logs all data sent to it.

Still trivially though for any talented reverse engineer. Somewhere in the code they have a function that checks if they think they're 18 and returns a boolean. Change the function to always return true. It would be harder if it was sending the image up to the server to analyze, but local is easy to break.