A fundamental law of the entertainment industry is that entertainment takes time to consume. It isn't like food or cars, where people buy to waste.

So no matter how quickly you can make a movie, it still takes an 80-100 or whatever minutes to watch it.

So they need more consumers. This is where agents come in.

They just need to convince people to want to own a series of robots to be entertained for them. I'm sure they could come up with a "vibe-viewing" UI to managing your little virtual popcorn-eaters - maybe Nintendo could help, they're good at that sort of thing.

And then kids could humiliate each other over how much media their peers are wasting without experiencing.

That's the smell of innovation.

I guess it is in the eye of the beholder, but "severe" seems overstated. I actually prefer some of the "restrictions" - no more Facetime spammers.

They should have known the proper way to do this is to create a Pinboard of photos of the sacked employees.

or I want to stop hearing the term.

Nope.

It will probably turn them into an also-ran, however, as the Epstein class congeals around Vance.

I think that means "when there is any alternative to Comcast available in a given area".

I used them for several years, and they dropped out nearly every time it rained. Which wasn't just an inconvenience - I work from home.

I got a second ISP account with a local provider, who didn't have the best reputation for reliability, but I used them as a backup.

After I set up monitoring on the lines I realized my connection to the local ISP was actually really stable, and dropped Comcast.

It was a great decision. No calling to negotiate the price down every year, it works in the rain, and the people there are really nice. They cost about 1/3 what Comcast charged, too.

If you're running telnetd in 2026 on a network with a gateway to the public internet, you have zero interest in security.

75% of the responses to this article are now off-topic.

See, agreeing to be subjected to ads is just payment for accessing whatever you're looking at.

So it stands to reason, that should apply to robots, too.

So wait for Google to start offering advertisers "AI ads", where you can poison someone's agent context for a fee.

"But only relevant context poisoning."

Because the author wrote it for the Mac.

You can run the gateway on anything.

Since World War 2, the US has been the destination of choice for most of the world's ambitious college students. I hope I'm wrong, but I think that's over.

Incoming international student numbers are down like a third, and dropping. Top schools in other countries with hiring budgets are gleeful. Once the current shifts, it doesn't come back - smart people keep going to the institutions where the smart people are.

The US is intentionally chasing folks off. If you take the National Security Strategy document seriously, the Trump admin is intent on shrinking the US from a superpower to a regional bully.

It will take a while to wind down - it was a really good thing. But long-term, say goodbye to being on top of the technical and economic heap.

The other part is there is a pot of money sitting in a perennially almost-useless company, and that is literally irresistible to a certain kind of person. (Usually, the kind of person who uses phrases like "unlocking value" to mean layoffs.)

Pretty sure a small pile of "ai ethics" projects could burn through a nice big chunk of that pretty quickly. And if 5% ended up in the right pockets, those pockets are set for life.

I work for a similarly-sized moneycorp.

I guess we're less top-down than some peers, at least about this stuff.

But we have a mostly free hand to use LLMs, or not. There has been vague encouragement to experiment with them, and the data security policy of course applies, but otherwise, no mandates or even heavy-handed suggestions.

I think the main use here is as a coding assistant, but engineers are expected to support/talk about/defend the code they check in, and the way we work enforces that.

We're building a robot to help with incident response and operational issues, and that's at the "vaguely nice to have" stage. It is capable of pre-populating context for a new incident in various ways - surfacing other recent incidents with the same application, listing relevant tickets and commits, and summarizing logs and instrumentation. All more or less before a human can get to the new channel. And it usually gets most of that right. But as far as actual troubleshooting, the thing is still tripping balls.

Some of that may be because old timers are still around - many of us have been here since the company was a nobody startup, and are the only available domain experts on how things work. We're also performing significantly better than the rest of the extended company, so they tend to accept it when we push back on something. Once the people who built this place are gone and the rest don't have that demonstrated authority, it will probably change.

So that's how Mozilla folds - bleeding off the reserves as part of the AI scam.

Wonder who will end up with the domain name.

You do realize they are the ones being tapped, right?

And would have to have the mental faculties to not give away the game hypothesized by the GP?

Loud and stupid is no way to go through life.