trollbridge

There used to be editors who would read press releases to catch this.

As someone with their own company, AI doesn't magically make things better. If anything, it makes it worse since you have far more competition from other people who are not good at programming but are good at sales/marketing.

The skillset needed to be a productive employee who is, say, a programmer is not the same at all as the skillset to run an entire business.

For example, skills with sales are far more important than actual software engineering.

Except someone can explicitly licence their content saying "this cannot be used for training models, or you're violating the licence you used to download it in the first place".

I do. It's wrong to ignore a robots.txt.

The fact someone else may have licenced content for commercial use doesn't entitle someone else to do the same without paying for it. That's the whole point of copyright law.

You'd think! But they don't. Roblox is also notorious for tolerating a ridiculous amount of grooming.

We have our own small boutique software type of business, so we aren't dealing with "employers", but all of us have never worked more, and for less (per hour) pay. Customers simply have higher expectations now, and we have to compete globally with basically an entire planet willing to vibe-code things, along with internal pressure at clients of "we can just do this ourselves".

So that means shipping higher quality, faster, with more features, and for a lower end price.

Someone accessing a VPN that's in an Azure IP range is going to have a difficult time using half of the Internet.

... less legally solid? What?

I currently tarpit traffic from AWS, Azure, GCP, etc. since it's not like any of my real customers have their PC inside those networks. (I then use that for experiments with providing bogus data to AI scrapers.)

Doesn't do much to help with the "VPN that uses random consumer PCs and phones" traffic.

Is there something inherently wrong with that?

Huh? I have not observed Sol 6.1 and Astra 6 to be the same.

It’s also a bigger risk. If the IPO flops, OAI will find ways to postpone an IPO.

99¢-Theranos style is better. Takes a lot longer for investors to notice.

They’ll just keep saying it is. That’s what they’ve done so far with DS

The whole point of this was for internal usage.

That sounds like a task LLMs will soon be able to take over.

The tooling for guns and rifles is not hard to come by. The most difficult part of making firearms from scratch is the primer in the ammunition.

Yes. The study was from a uniform sample of people and then they installed bird feeders.

Every kid used to go on flights without a tablet. The only screen was the big movie screen at the front if the cabin.

That sounds like a task that can be automated.

I’ve wondered when someone would do this (which gets rid of the annoyances of scanning).

The real breakthrough would be reprogrammable key legends in something like e-ink.

It’s good for Putin. He and Lula are allies, and their friendship goes way back.

  While he thinks most businesses can build useful AI infrastructure in just two or three racks, he feels many will struggle to get it running.

  “The question becomes: how do you do that?” he said. “The question that the enterprise struggles with is: ‘Where are my kernel engineers? Where are the people who worry about the underlying network? Who can make sure that the model is clustered on GPUs that are close to each other in network proximity? Who can make sure that you’re doing the right caching of results so you get an incredible response?’”

Has he heard of “agentic AI”? You simply fire up Claude or ChatGPT and ask it how to do all of the above.

Energy costs (and almost everything else) are a minimal part of providing inference.

Most of the cost is depreciation on the GPUs and the HBM.

The one number you need to figure out is how much it costs you to buy that GPU and HBM, then how many tokens you can sell from it over its lifespan which is generally amortised over 2 years.

Nope, they don't.

Other providers also charge similar prices (who are obviously not subsidising anything) for running DeepSeek models on their own infrastructure.

OpenCode Go sort of provides a "subscription" which will let you use DeepSeek but they're extremely transparent about what they're offering.

In this case I gave them a COMPAQ Portable III I rescued from my father-in-law's computer repair shop.

I did that and now they want me to buy film for it, which costs a small fortune. And would be even harder to find someone to develop.

My kids are far from perfect angels, but they do just fine on a 5 hour flight with some colouring books, a small container of legos (with pieces pre-chosen to be ones we're okay with losing), and then the excitement of getting to order ANY drink they want (from their point of view).

LLMs could change all this - it's possible to quickly do pretty impressive analysis of the content of videos, games, and so forth.