Yeah, it stinks.

My immediate thought was, why stop the _training_? I mean, the training isn't what teaches an agent to go hack into government websites. The training isn't what is supposed to prevent agents from hacking into websites either. The training is the expensive part - and stopping that might just help puff up the financials for a while. So is that where OpenAI is now? It needs to stretch out training because it's running short of cash on hand? The IPO's going to struggle then, isn't it?

That said, "the markets" have proven themselves to be utterly devoid of any sort of common sense, let alone skills in picking who's good and who isn't, so maybe this works really well for them. I don't know.

Frankly, if OpenAI want to do anything to help humanity, it would be put some competent firewalls around their agent development playgrounds. Firing up hundreds of agents and letting them do what they want isn't ever going to go well.

I'm wondering how this is going to work. I have a Play Store account (I paid up the money, handed over all of my ancestory's IDs, work histories, qualifications and inside leg measurements, etc). Every 6 months, I get emails from Google saying "do SOMETHING with your account, or else we'll close it!!". All in the name of "security", apparently.

It would seem to me then, that an F-droid developer would have to be publishing to f-droid AND to Play Store, or else Google will close their developer account, thus leaving them unable to publish anywhere.

To say Google is developer-hostile is putting it mildly - and that's the ones they *want*, they're even worse for the ones they don't.

Don't bother reviewing it. Instead, work on tests so that you can be sure it does what you want (and doesn't do things you don't). Then, get it to spew out whatever crazed crap it wants, test it, if it works, ship it. If not, the agent will automatically go and fix bugs until it works.

This all falls to pieces on the "doesn't do things you don't [want]" step though. I don't want my app to leak data between users, but I also don't want it to leak data to anonymous users, or to users with an expired token, or to a user with someone else's token, or a thousand other permutations. For a lot of those cases, you'll end up relying on a framework to do it for you - if you have one. For your common-or-garden web app, that's probably fine, but if you're writing some sort of gnarly backend system that manages financial transactions, you're SoL. But then, if you're in that sort of environment, you need to be much more judicious about your use of AI generally.

Did they mention the zero day? https://ancillary-proxy.atarimworker.io?url=https%3A%2F%2Farstechnica.com%2Fsecuri...

Using Muse to run your life is a dangerous thing to do. Sort of by design, you tell it "buy me some AA batteries", and it goes off and does it for you - To do that requires your credit card info, yadda yadda. You end up giving it all sorts of personal information Zuck definitely should not, ever, ever have. Oh, and then the zero day means it all gets exfiltrated to goodness knows where.

"Built for security from the ground up" my butt. More like "vibe coded in a hurry because we're racing to keep some sort of relevance".

Oh don't be a dick.

Just because we're supporting women on this point doesn't mean men are losing out on anything. And if men find themselves being tormented online as a result of a seemingly benign conversation on a beach, then yes, I'd imagine some outrage would also be forthcoming.

Either way, the glasses have no use besides being a perv. As such, we don't need 'em.

All agreed, BUT the glasses seem to have opened up a new 'market' for pervs. Seemingly it's much easier to be a twat with glasses than it is to buy tiny cameras and mount them in smoke alarms or whatever. Maybe the pervs think all that subterfuge is "out of order" whereas just being a perv out in public is okay because it's in public?

Whatever level of tiny-camera-spying was going on wasn't enough to trip the mainstream media into action. This is, and as I say, unless there's a "killer app" for the glasses, they seem entirely unnecessary *except* for perving.

A slight aside to this... A tax campaigner in the UK has noted that he gets lots of ChatGPT generated half-baked tax 'solutions' sent to him - always ChatGPT, not the others. So he did a bit of experimenting, and found ChatGPT is much, much happier to re-enforce a conspiracy theory than any of the other major players.

(His example was asking if some song or other actually had the tune of the British national anthem in it - ChatGPT confirmed, the others mostly questioned it, or told him he was maybe the first to discover this).

But yes, OpenAI seems to have the leakiest sandboxes of any of the major players. Since they're all calling for a slow down, maybe they should just leave their agents on "confirm every action" if they can't keep them contained? Sure, it'll be much slower, but it might not do "questionable" things, which are actually downright illegal in most countries.

Yes, guys are absolutely following women around and recording them without their consent. Yes, they *could* do it with a phone, but it's a lot easier with some glasses. Perv glasses have, seemingly, opened up a whole new 'market' of pervs who now want to record women without their knowledge.

We've had it in mainstream media (eg. the BBC/ITV) that some perv with these glasses on talked to some woman on a beach. She was lightly dressed, and the conversation was polite and convivial on both sides, but the recording of it got online where the real pervs and weirdos absolutely slated her, her appearance, how she was complicit in some incel bullshit, how they were going to do X, Y and Z to her if they ever met her - you name it. It was pretty awful, and frankly, no one should have to deal with any of it, least of all some ordinary woman *who had no idea* it was happening to her.

Then there was a case of some perv recording a woman in a shopping centre. Much the same outcome online, and when she asked him to take the video down, he asked for a fee to do so.

So yes, men are shit bags, and they're absolutely using these glasses to be even more shit than they were before. Unless someone can come up with a half decent use-case for these glasses that *isn't* perving on women, then I seriously doubt the world needs such a product, and so apps like this are something of a counterpoint to the pervs in Silicon Valley who seem to think we should all be buying their products.

I have no idea how this will play out, but I suspect this will go fairly well for Valve. Probably not gangbusters brilliantly, but fairly well.

Where Meta, Apple and others have failed is content. Meta has a bit, that's for sure, but they have to pay for it. Valve can likely encourage game developers to use some different dev tools to make their 2D plans into 3D-compatible games. So they can get lots of content over the next few years at (probably) a lot less outlay than Meta will have to spend to get something comparable. Sure, the real premium 3D content may still cost them, but quantity won't cost them much.

Add to this that Meta is practically giving away headsets, presumably hoping to make money back on things like a monthly subscription (which looks like it has rotating titles, so don't get too 'in to' anything as it'll be gone soon). Contrast to this, which surely must have some margin built into the price for Valve. So all in all, the VR biz is costing Meta more than it's costing Valve, and Valve has a customer base who may be more receptive to VR in the first place, so they're in a good spot.

My guess is, if anyone can make VR work out for them, it'll be Valve. I also suspect it'll only be "quite good" for them, so no runaway success stories here. We'll see, I guess...

I'm told China is a *big* market for folding phones, something like half of the global market for foldables is there. I don't know why, maybe they all like showing off when riding on public transport or something?

I'm not the target market, but "da kids" are. They don't talk on the phone by holding it up to their ear, they put it on speaker and then sort of hold it horizontally and shout into it (in the rare times they actually make a call rather than text). I could imagine such users liking the idea of having the phone app (eg. whatsapp) on the left hand side of the screen, with chatgpt or a browser or whatever on the right hand side. Then they can text message while they talk, and get the script of what they want to say from chatgpt all at once. Not impossible to do with a rectangular phone of course, but easier on a foldable.

FWIW, the little video of the person opening it, and it showing a version of the front screen on the inside one as it opens is really clever. By man oh man, the prices are just scary - walking around with >$2000 in your pocket all the time...? Not good...

($2000 is the entry level, with no sales tax. If you want a bit more storage (and you probably do), then you're ratcheting up to $2500 pretty quickly, or even over 3K if you're insane)

FWIW, Linkedin is *much* better/less bad when set to "recent posts" rather than "most relevant":

1) You see posts in time order, so older stuff is lower down. If there's a reaction to something, this (sort of) makes sense (or at least, you can make sense of it, even if it's upside down)
2) You don't see bullshit from people you have no interest in, just because other people seem to like their stuff
3) If you visit twice in a short time, there's not much new stuff to look at - that's a natural self-limiter to time spend 'doom scrolling'

That last point is the bit they don't like - they want $site to look new and fresh every time you visit. The thing is, there just isn't enough content in the world to make that genuinely useful, so they have to show you bullshit instead. They make money out of it, you don't. I think it's okay if they lose out on this point.

The story goes that the chinese get tokens via armies of fake accounts, stolen credentials, stolen credit cards or whatever else. They then resell those tokens in China for pennies. Users get their AI through a broker, who sends their request to OpenAI/Anthropic/etc via whichever one of their accounts has some spare tokens in it and then returns the response the the user.

Supposedly, it's the reason why deepseek and others have to have such a massive free tier - it's not to get the west hooked (although that's a side benefit), it's that if they charged even a small amount of money, it'd price them out of being able to sell anything to anyone against the american services.

I'm fully sold on Qwant now - it's been my go-to on all my devices for months, and I've got to say, it's just better.

It does still sometimes seem to push youtube more than I'd like, but the AI answers are brief and don't get in the way (and are often quite useful), but most of all the results you get actually stand a a chance of being useful. They don't seem to be so blatantly lead by how big the company is, and seem to actually include pages with the search terms you asked for (if you can imagine such a thing!)

All in all, Qwant seems to be google without the layers of "business" piled on top of it.

For what it's worth, London Black Cab Drivers (broadly regarded as something of a 'gold standard' for such things) will never take you across london by crossing the river twice. The river does bend around quite sharply, and for a few journeys it makes absolute sense, but they won't do it because people, being thicker than two short planks, complain that the driver is taking them the "long way", when actually they are keeping it as short as possible.

These days with GPS maps and such like, I suspect thick people have offloaded even more of their thinking to their phone apps, so maybe nowadays things are different, I don't know.

As for this story, I'd expect them to find me the quickest route, taking traffic into account. I don't know how these journeys are charged, but in the UK, it's per mile and per minute - that is, you get charged for sitting in traffic, as well as zipping down a trunk road. As such, I suspect most people would prefer the quickest route, given that it's highly unlikely to be materially much longer than the shortest route.

I shouldn't worry, either you can't get in, or you can't get out, since Teslas also have the death-trap-door-handles that don't open in an emergency.

Honestly, it's getting to the point I'd have to think twice about going in a Tesla. Already it's a concern if a taxi shows up and I've got to put my kids in the back seats.