Sounds like a great argument to turn these things off

"Why did the probabilistic text generator fail to exactly follow the constraints?"

Truly a mystery. Better buy more SaaS to smear on top

Fair point. Probabilistic systems being unpredictable isn't news. The article's just noting current observability tools weren't designed for these failure modes.

Offtopic: the "scan line" thing that the left-hand pane labeled "Documentation" is doing takes a continuous ~12% of a CPU on a Mac Studio. Incredible waste for a pointless visual effect.

    Since privacy will be an issue, "Shazam-like" filters will inhibit uncleared capture of voice.

So now the operating system will decide which recordings are "cleared" and which aren't? Fuck outta here with that nonsense

Reminded me of the recent Q.ai acquisition by Apple. It's obvious that if you're using voice it should be as clear as possible.

I was going to suggest the next big leap will be some kind of "OmniLinux" that spreads across all devices, appliances and hardware that contains any kind of OS and enables interoperability, control and telemetry. Allows updating firmware from a central point, access control and power management. Will be used by humans first, then bots later. There might be some big retro movement to old world things as a result when people reject the idea of a "common dashboard" for the things they own. Might be useful for sharing and rentals though. Is a new OS needed for this though, why not some standards and protocols?

Had this vague thought that the OP is a bot. Does it matter?

OP is not a bot. I can tell you that for sure.

To your point, and I see that it is a recurring opinion around here.

I like the way you think, yet I find it difficult to see how one such movement would emerge.

Most likely from the Open Source community, as I do not see any incumbent intending to go in that direction, don't you think?

I see it as a rather logical step with the advances in voice first AI wearables.

Think about it. Not everyone wants to be recorded as a bystander. Privacy will be an issue.

The technology for audio signature already exists and works fine.

It will be a matter of opt-in/opt-out from users, not an OS decision.

This isn't science - it's slop

    Also, since almost no one writes tests by hand anymore and
    instead generates them with LLMs

For suitably-small values of "everyone", I suppose

Mmmmm tasty slop

    use AI for code review, having another AI (or the same AI in a
    new session) audit code. It can examine every line carefully,
    never tires, never misses.

"never"? Absolutely roflcoptering over here.

If you've never had an LLM code reviewer hallucinate, you're either not reading the output or you're just lying.

I am shocked, SHOCKED I TELL YOU, to find out that executives might be lying!

Me too, I was very loyal and then got laid off.