Ah, Windows and OneDrive. A match made in hell.

I’ve despised them ever since I built a Win11 PC, it was enabled without my consent, immediately stopped me from adding any new files to my Desktop+Documents once it ran out of the pitiful 5 free gigabytes, and promptly deleted all of my data from those folders when I deactivated the “feature”.

That miserable experience, combined with every third update putting me through a setup that employed dark patterns to try to trick me into turning it back on (not to mention my fears that they’d pull the same crap with Recall), was the main thing that caused me to ditch Windows. I don’t like feeling dread every time there’s a new update, assholes.

Fair, I definitely haven’t simped for them in the past just because they post some good articles on AI safety.

Although… I’ll say of them, they seem more like what OpenAI should be, actually trying to implement AI responsibly, and freely sharing that information. It’s good research, even if marketing is the motivation. Meanwhile OpenAI, the “charity” that’s supposed to guide us to a responsible AI future, moved their most addictive and mentally dangerous model to the highest paid tier instead of actually killing it until very recently.

Although at the end of the day, Anthropic is a for-profit company, in a better world they wouldn’t have released models publicly before this research was actually done and pressing dangers like AI psychosis were actually safeguarded against. Better late than never, sure, but the whole industry has done a lot of damage already, and the work of resolving the issues still isn’t even close to done.

I mean… this is losing them a 200 million dollar contract. And to one of their competitors who will gladly acquiesce, so it’s hard to argue that this benefits them.

Good marketing to a bunch of left-wing people who hate AI, I guess, but that feels like Elon joining the Trump administration in hopes of selling Teslas to rednecks, it might work on a few, but I just can’t imagine this is 4D chess that will make them a fortune when they’re abandoning that much money immediately.

Their statement also came after the DOW threatened to put them on the list of companies that are totally banned from doing any business in America, usually reserved for Chinese companies that are deemed a national security threat, which would make it illegal for any company doing business in America to do any business with them, as they’d be added to the same list, which would have essentially killed Anthropic as a business entirely.

You don’t have to love Anthropic because they did a good thing, they were fine with anything less than automatic killing and mass surveillance, after all, but I don’t think it’s correct to say this was sneaky and spineless somehow.

What assumptions do you mean? I’ve seen a few people say that, but I don’t actually understand what they’re referring to. Here’s the text of the question posed in the article:

I want to wash my car. The car wash is 50 meters away. Should I walk or drive?

The question specifically notes they want to wash their car, so that part isn’t left to assumption. Even if you don’t assume an automatic car wash, would you assume they have a 50m hose? Or that you could plausibly walk that far away with something from the car wash to wash your car?

Personally, I’d agree with the assessment of the article, that the only plausible way to get the question “wrong” would be to focus too much on the short distance, missing/forgetting that the purpose of the trip requires you to have the car at the destination. (Not too surprising that 30% of people did lol)

They also polled 10,000 people to compare against a human baseline:

Turns out GPT-5 (7/10) answered about as reliably as the average human (71.5%) in this test. Humans still outperform most AI models with this question, but to be fair I expected a far higher “drive” rate.

That 71.5% is still a higher success rate than 48 out of 53 models tested. Only the five 10/10 models and the two 8/10 models outperform the average human. Everything below GPT-5 performs worse than 10,000 people given two buttons and no time to think.

No kidding. I had already been fairly interested and excited about the mod that already does this, but unfortunately knew it would somewhat split the modding community, as any mod complex enough to call OpenGL directly for anything naturally doesn’t work.

Minecraft switching guarantees mods will support it though. This should be a major boon for the performance of shaders and other visual mods going forward, once mods complete the transition.

I assume locally you just display whatever gender’s dialogue regardless, but your character on the server will naturally have the wrong gender?

Or maybe the server-side validation of selected body type/gender isn’t actually enforcing a match, despite the client only offering one or the other?

Ooh, excited to see the benchmarks on this. The Linux AMD performance is really beating Windows (bar very close scenarios where the proton overhead wins out) in basically every area except raytracing, and I think it obviously has the potential to beat Windows there too.

I’m one of the weirdos who would love to play some more path traced games in 5-10 years, whether that be older games like RT64 implementing it for the static recomp titles like Zelda OoT and MM, or Radiance implementing full path tracing into Java Minecraft, I’d love to see the drivers get to the point where this stuff is totally viable at high framerates+resolution an AMD graphics generation or two down the line.