With this character’s death, the thread of prophecy has been severed. Reload to restore the weave of fate, or persist in this doomed world you have created.

Microsoft going out of business/doing severe restructuring or downsizing

Although I wonder if they could. Microsoft seems like one of those “too big to fail” companies, where they’d never be allowed to fall on their face, since Azure and Exchange prop up so many things. It’s not like there’s a major second option for an OS if you just buy a computer off the shelf like a lot of people do. You either get a Windows or a Mac.

If that’s the appetizer, how juicy’s the entree gonna be?

At the risk of going on a tangent, isn’t the entrée the appetiser? You don’t have an appetiser, an entrée, and then the main course.

It’s also a big attack surface. Just like how a lot of malware looks for the browser password cache now, it doesn’t take much for a malware developer to just go for the recall store. The malware doesn’t need to pack in software to take screenshots, if the OS serves it up for them on a platter.

The year of the word processor approaches

Also to rouse and inspire them.

“I’m working on a machine that will make you guys redundant, and then I’ll make those booing me see. l’ll make you all see.” is hardly going to do that.

He would be much better off talking about how it was the stuff of science fiction not long ago, and how the graduates would be helping to push humanity forward, and make real, things that were also previously considered impossible.

Some of the talks are also just really bad. I’ve seen a few, and they’re little more than ads, or bragging about a thing the institution is doing that’s unrelated to the graduates themselves. Saw one where the speaker was talking about how the college was using AI for various things. Why even have that in the graduates’ speech?

Distillation isn’t stealing the original model, though. It just uses the models to make synthetic training data to train their own thing. They aren’t stealing the model itself.

Plus, a lot of companies do it. Anthropic’s Claude was calling itself DeepSeek for a while.

It also doesn’t seem like as big a deal as Anthropic and Open AI make it look, IMO. Them treating it like a national security issue where the company gets its models stolen from under its nose just comes across like a media company claiming that every download is a copy they would otherwise have sold at full price, and thus they have accrued trillions of dollars in damages.

I could, in theory, take a bunch of google Gemini outputs, and train a GPT-2 model on them. That doesn’t mean that I’ve recreated Gemini, nor does it mean that i’ve stolen it from Google, either.

To top it all off, it’s not like their services were abused. The companies were presumably paid appropriately for the usage.

Their reputation is also a bit in the toilet, because people hear “AI” and think of ChatGPT.

So “man hospitalised after AI suggested he put glue on pizza for tackiness” would have people think he was using it, when he might well have been using a different LLM.

The cost has also shot up because a lot of the new frameworks are much more token heavy than the old ones.

So the original free plan might have made sense when people were only typing little questions into it, and using a handful of tokens, but is no longer cost-effective with things like modern agent pipelines constantly throwing tens of thousands of tokens at the service.

I tried running a little locally hosted agent thing on my computer the other day, and it was feeding a hundred thousand tokens at the model every few minutes, because it was keeping all the files in context. Sure, it hit the cache a lot, and so the effective cost would be less, but it’s still a lot more token usage than me poking the model with inane questions.

Do they even make enough heat for that to be viable option? Most computer systems can handle a pretty low temperature before they start having problems because they’re over-heating.

I don’t know, it has the opposite effect, IMO.

It just makes them seem obnoxious, since the example they chose was a parent who was distracted with the computer open in the changing room while they were supposed to be helping their children with their skates, and literally mentions how the other parents have to navigate around the thing.

You’d be more inclined to think that they’re a computer addict who can’t put the the thing down for even a moment.

On top of that, the video is basically a recipe to drop the laptop and have it shatter into fine powder, if you’re holding it by the corner like that.

As many as they need to get a 51st 51st state.

oh no, you don’t want to do that. The adhesive is an anti-nutrient, you see.

Completely neutralises the fibre. You’re better off snacking on some blotting paper.

I’d be curious if there might also be a cultural aspect at play.

Apparently in America, their portions tend to be quite large, since the expectation is to get as much for your money as possible. Anything you can’t stomach can then be taken home to finish another day.

Whereas many other places don’t tend to do that. Food served in the restaurant is to be eaten there, and wanting a take-away container to take your meal home means paying extra for the container.

They’re used for some trains now, though I think that a lot of them have since switched to rheostat or regenerative braking instead.

Slightly odd choice to use a motor instead of an eddy current brake or some such, when it’s supposed to be a drop-in replacement for existing braking systems.

Is it supposed to be a quick hybrid conversion system rather than just a brake?

EDIT: I’m not sure if it is. The article makes it unclear, but going by the manufacturer’s site, the electric motors are meant to replace the piston on the caliper, rather than using the motor itself as a brake.

It’s still a mostly conventional braking system.

We had a rather nice thing going with pure HTML. Sure, it wasn’t the prettiest thing, even with CSS, but almost every device could run and display it in its own way.

You didn’t need a custom thing, or a bunch of extra code adjusting the webpage for each type of device that opened the web page, since that job was all done by the browser.

That’s basically model routing, and has existed a while. Open AI’s GPT-5 and llama-swap do that, for example. If the task is simple, it uses a smaller, less intensive model, and only uses the slower, larger one of the task is more complex.

Though most tend to operate with models on the same device/service, rather than a model run elsewhere.

Back in my day, computer was a job, and all you had was an abacus. We liked it that way. None of this newfangled al-gebra nonsense.

It’s also cheaper, if they can offload a portion to the user’s computer.

Sidewalking