That's the in-house preprocessing hardware, not what they're training on.

Yes!

It's still a weird article, to highlight a "big" storage appliance. Having all that NVMe local feels like it would be much much much much faster.

That's exactly what European governments and corporations will have to start doing. Adding the US to the same list as Russia, China, Israel, Iran etc.

Agreed. As Simon Willison points out, November 2025 was a a critical months because that's pretty much when coding agents became «good enough», eliminating most of the problems pointed out in this study.

I regularly see Claude Opus 4.7 dropping constraints from an otherwise small CLAUDE.md at merely 20% context use. I have to keep reminding it, and it has all info ready in its context, still time to time decides to ignore parts.

GPT-5.2 was released after November 2025.

The post is written like as if consumer coops is a radically new idea (spoiler: they're everywhere in Europe). Unfortunately they have long since succumbed to the same mechanisms as their competitors and have a tremendous power concentration in top leadership, with directors making ridiculous salaries. On paper though, e.g. Coop with a 25 % market share in Norway is owned and controlled by about 1 million of its customers.

This idea is model after those consumer coops from JCCU to mondragoon to some other smaller american consumer coops. Our proposition is that the openly displayed set of the CO-OPs goals along side AI moderation may help with alleviating these issues

Hah, so this. USSR had it formalized in form of PotrebCoop, literally consumer cooperative in 1921.

Consumer cooperatives go back to at least 1844

https://en.wikipedia.org/wiki/Rochdale_Society_of_Equitable_...

By now it should be common knowledge that telling an LLM not to do something is not a «safeguard». Access control is.

This looks like some kind of marketing. Also, the equivalent of spec work. The NDA/secrecy also means any time spent on this is completely meaningless to the participants unless they win the lottery, because results can't be published.

It looks like if they reject paying you any bounty you would you still be bound by the NDA. If so, then they could both not pay you and still spike the story. That’s not something I would ever agree to.

I know we celebrate cynacism online, but $25k to OAI is like $0.025 to you and me.

Skimping out on 2.5 pennies you promised someone is cartoon villain levels of greed.

Yes, I know, Altman is a cartoon villain. But please, they are spending more money decorating their bathrooms. They'll pay out.

Will they? The NDA makes it so if they don't, we'd never know. Bug bounty programs suck but they're better than the alternative, but even running one openly, there's always convention about whether the bugs being submitted are real or not, with a lot of low quality reports that the submitter thinks are gold. That happens out in the open. Now add an NDA into the mix. Sam's reputation doesn't even have to enter into the equation for it to be a bad deal.

Surely it is marketing. It’s some “we are danger” narrative, from Anthropic Mythos and now OpenAI too.

OpenAI was doing this back with GPT2, saying it was too dangerous to release

Dario said the same thing about GPT2 when he was at OpenAI. As you can see the digital and physically worlds are now completely compromised and life is a pale shadow of what it was 5 years ago…

These guys have poor track records and compromised incentives.

I have no problems using my AirPods across two Macs, an iPhone and Windows. I have to manually reconnect on Windows if I have an active Apple device nearby which I recently used the AirPods with, but apart from that it's quite seamless. This worked fine in 2020 already.

How is having to manually reconnect each time when switching between, say, an IPhone and a Windows laptop every time fine or even remotely seamless?

I do not have to that even with my ~30€ JBL buds.

I've seen A LOT of public sector projects starting out with loads of Azure services and >$3000 montly bills for applications that could've easily run on a single VM.

I've seen Big Data pipelines (Hive tables, Spark jobs for queries, data engineers setting it all up) for what was ultimately a 5-10 GB dataset.

Companies cargo culting Big Data stacks when their data is nowhere near big is very prevalent.

This a structural problem not an awareness one. Is not like they don’t know they can run it on a 5$ VPS, the problem is that there are no incentives to do so. You’d be surprised to know how much of engineering is there to address organisational challenges rather than technical ones (ie: micro-services)

Same. They've been sold a bill of goods by big cloud tech and company.

Yea, that's ugly. I'm sure it could've been done more gracefully with 15 minutes more effort. But judging from the general wear and tear on this poor Mac I guess they don't even consider the resale value.

I can't even imagine prioritizing resale value here over one's own comfort. The purpose of a tool is to be used, not to serve as an asset class.

Sure, but comfort != abuse :D Apart from the filing, I can't think of ways to make such a recent Mac look like this. Did it suffer a plane crash? Acid attack? Thermite fire?

I appreciate the customization, but would probably make an effort to make it not look like (another) accident.

I think he is not worried about the resale value.

> This was on my work computer. I expect to similarly modify future work computers, and I would be happy to help you modify yours if you need a little encouragement.

I don’t understand the actual decision but I appreciate the gusto with which it was made.

The main reason to consider resale value is 1-2 years later you may want to upgrade and selling it to another person typically yields you more money than trading it in with Apple. Doing something like this may decrease how much you could sell it for later.

If you’re not planning on doing that then it’s not really a factor for you.

> work computer

Seriously, I have several mac laptops dating back to 2004 and they all have less wear than that.