0x3f

Yeah, thats explicitly what I'm saying so I'm not sure it needs repeating. That has very little to do with it being lazy though, is the point.

We were already implicitly discussing RoI when we were talking about 'legal consequences' above. This is how people decide between alternatives, generally.

Every moment you spend doing accessibility is a moment you spend not doing other things. You could argue it has a high RoI to do accessibility, fine, but that doesn't make it lazy _not_ to do it. Maybe I have even higher RoI/EV stuff to be doing.

It's much more verbose and can't do everything Tailwind can anyway.

E.g. how do you style a child on parent hover with the style attribute?

Sounds like you're kind of just talking your book though. Person who makes accessible sites suggests you need an accessible site. Blind people aren't the only ones who might need modifications. You could have an infinitely long list of adjustments for all kinds of disabilities, and tell me I'm lazy for not doing each of them. Why are blind people special?

Perhaps, but that doesn't make a win for the copyright lobby a win for the little guy. If anything, pro-copyright decisions are decidedly _bad_ for the little guy.

They might 'win' reliably, but I don't believe the margin on outcome itself to be that great.

But in order for that to be required, it couldn't be 'blatant'. Unless you mean the admission itself is 'blatant'? Then, I suppose so, although the normal interpretation of that would be from the content itself.

HN really loves the copyright lobby when it's against someone they hate, huh

> A one shot prompt can only get an estimation of what you asked for, no matter how descriptive you are.

Isn't that true of all creative outputs, even of 'legacy' provenance?

> Even at worst, it expresses intent, which has meaning to humans. We are humans. I can't force you to do anything, but I can ask you to. Don't disparage what it means to be humans talking to each other - it's one of the few things we have left on Earth.

Isn't that what... signalling is?

Well what purpose does the rule serve if it can't be enforced, if not signalling/norming?

You have to consider that any AI content worthy of the Oscar shortlist is going to be very high-quality, and likely intensely hand/human-tweaked in the first place. It's not from the general population of all AI content out there.

> I have always heard that dopers are consistently ahead of testing regimes

I don't know about that, even the very biggest names with the most funding quite often get dinged for it. I suppose I'm not really saying that the detection rate for doping is high, though, just that it's much higher than AI detection in high-quality content (which I would suggest is approximately zero).

I don't think AI-generated 'avatars' are anywhere close to being Oscar-worthy as things stand, so it seems kind of a moot point (hence the 'signalling' thing).

If they ever get that good, I would just say you can't really fight the market. If AI content is good enough that people want it, then the Oscars just get left behind after a while. But that's fine, and up to them.

> Also, "truth" is a thing that exists, and just because you can't always tell if somebody cheated the rules or not, does not mean the rules are "performative signalling".

I don't really understand. If you can't hope to discover the truth, in what way is it not performative or signalling?

I'm not sure how true this is.

If you consider low-stakes crimes, typically to get to a steady state of effectiveness you need at least some sort of bootstrapped period of ubiquitous enforcement. If that's impossible then I'm not sure you ever get to effectiveness.

If we're talking high-stakes, death-penalty-lottery-if-you-break-the-rules type stuff, then I think actually detection rate (i.e. consistent enforcement) is the biggest predictor of reduced rates, not severity of punishment.

Can you explain how an Oscar-worthy piece of writing would somehow be able to contain blatant AI-generated content? How would it have already passed the good-enough-for-an-Oscar filter?

How are there consequences for being caught if it's impossible to detect?

Moreover, why stop here? There are many great rules that are impossible to enforce. Why not a rule that the author isn't allowed to have any racist thoughts when writing the material?

We can't read minds, but it sure is a nice thing to care about, don't you think?

It's much easier to tell if athletes are doping than to 'detect' AI in text that's already Oscar-for-writing level good. I would suggest the latter is quite literally impossible.

Obviously just performative signalling that doesn't really do much. You can't definitively tell if AI was used, so the rule can never realistically be enforced.

Then again, the Oscars are surely almost entirely vibes based anyway. So it's hardly some internally consistent system of merit in the first place.

I think people are often just logged in for years and forget they're reading a paid resource.

Rent control is the canonical populist price control. That one's evergreen. Egg-prices-by-fiat are just a fad!

There is no way Walmart or Target or whoever is giving up their gold dust to some nameless SaaS for pricing. They will do it in-house. Never mind the fact that a similar aggregation attempt for the property rental market was considered actionable by the DoJ already.

Just seems like a difference of degree. You have n price tiers in both situations. Traditionally, the complexity of n_prices = n_customers (or even n_prices = n_customer_contexts) was too painful to be worth it. But they were always approximating this up until now. 'Categories' are just wider buckets over individualized prices.

Yes (after user protest as I recall) but then the new UI just diverges over time, and quite possibly gets features not in the old one. I want the latest UI, and ideally the default UI, I just also want that not to be stupid.

To be clear I don't even think I'm talking about taste here, although people did complain about that. I can't think of any good reason that 'fill' is not the default action on an app/extension whose core purpose is to fill things.

The Bitwarden UI/X changed relatively recently so that detail view is the default click action now, rather than fill. I don't think I've ever actually used the detail view, because the edit view does the same job.

Never mind that 'fill' is 100x more common as an action. So why on earth is that not the default? It is indeed an unfathomably stupid UI decision, beyond what I regularly see in other apps that I use.

I still hate it to this day, and find it incredibly clunky. In fact that alone is kind of making me want to give in and just use Apple.

Are you really concerned about asking these kinds of questions though? Like how many LLM-able Tiananmen Square questions are you needing answered per month really? And it seems like you know not to trust it, so there's not even a risk that you're going to ask such a question and rely on the answer.

I run into Claude being a stubborn idiot about far more useful stuff all the time. And often all it takes to bypass is starting a new chat and reframing it, so it's entirely pointless hand wringing.

Then let's not forget only one of these is a paid product, and it's not the more annoying one. I feel like I can forgive DeepSeek for just obeying the laws of the country they're based in, as silly as those might be, because they're being pretty generous with the weights in the first place.

I think you've just got perceptual tuning for the human to human context. It's like how human faces that are 5% different can be very obvious to us, but it's much harder to tell (for example) two cats apart at the same difference level. All of your adaptive pressures are pulling in that direction.

But at a system-scale view of intelligence, humans are squashed together and it plays a pretty small role on outcomes. You should much prefer being rich than smart if your goal is 'success' by most metrics.

The CV of e.g. IQ is only 15%. That's in line with other 'natural' attributes of humans, but not compared to something like family wealth or background. From a quick Google, wealth is 700% in the US? Income also same OOM.

Even if you take IQ, a 2-sigma event on that distribution is 'only' around 130. A 130 outpaces a 100, sure, but it's hardly full on dominance in most tasks.

Certainly much less powerful than just having rich or pushy parents.

> The CV of e.g. IQ is only 15%. That's in line with other 'natural' attributes of humans, but not compared to something like family wealth or background. From a quick Google, wealth is 700% in the US? Income also same OOM.

I suspect Demis 'gapped' you structurally (upbringing, early decisions, luck, positioning), moreso than via intelligence, which doesn't actually vary that widely in humans.

> more people getting scammed via "please tap the build number seven times and then go to extra settings and enable untrusted installs and then go to this website that I will dictate the URL of and you should ignore that install warning" etc etc etc.

I don't really understand. You seem to be against the 'annoyance' of the protections, but that annoyance is precisely why the scammed count is lower, no?

I certainly believe _more generally_ that the market for scam victims is much bigger than the market for sideloaders, for example.