camgunz

You aren’t thinking this through.

No need to be so polite; I might just be very stupid.

You were just told that this wasn’t stacking but you keep insisting it is.

Having an LLM generate something and then asking another LLM to generate that something is asking for 2 hallucinations, not 1. This doesn't go away the more LLMs you throw in there. The only reason this works at all for things like ChatGPT is their prompts are super different (and the editor LLM is way cheaper, of course).

And you still have to check for hallucinations (before you say throw another hallucinating LLM at it know I will make you pay infinite forum shame points if you do).

the same solutions that solve the Knight and Leveson issue for humans solves it for LLMs as well

No, LLMs are plagued by the same problems: https://arxiv.org/abs/2606.20158. After all what are you gonna do, say "build me a SQLite, but definitely not like [other model X] would"? The solution is a very unambiguous spec, but that asymptotically approaches code soooooo....

Finally, I think you're hand-waving NVX. A program that can digest any spec and measure how well any program meets it is pretty wild. Good luck with that one.

Reliable AI is an oxymoron and deeply misleading. Even if you can stack models the efficiency tanks. Also, just looking at the Wikipedia page (https://en.wikipedia.org/wiki/N-version_programming#:~:text=...), don't LLMs already have all these problems?

Upvoted for real math

no ur wrong

Error percentages compound. Having an LLM do something with a 5% hallucination rate, then having an LLM (same or different) with a 5% error rate check it means there's a 10% chance of error, not .25%. Prompting can't save you here; it's the fundamental math.

TFA says "no hallucinations" but you can't fix hallucinations with rules or other models. I know I'm screaming into the void but whatever.

First, I beg people to read Annie Lowrey's "The Great Affordability Crisis" (https://xcancel.com/i/article/2077113148524417439), which is criminally under read.

Second and more fundamentally, most people know neoliberalism fucked us. Like, not just that we're at the end of a typical cycle, but that what predated it was better. That's why the middle is falling out of western politics and populism is rising: voters are bitter, angry, and looking for the next thing. It turns out you can't turn everything--romance, friendships, caring for children, education, health care, free time--into a (preferably unregulated) market and still have a functioning society.

Third, we have to ban social media, and here I mean any communications platform where you can farm engagement and virality (Reddit is social media because there's a feed and engagement metrics; Usenet and forums are not). This seems like a wildly hot take, but the truth is we should ban video news; only banning social media is a huge concession.

Fourth and finally, don't fall into the "make America great again" trap. Pick a decade and I'll point out some insane awfulness: Jim Crow, marital rape wasn't a concept until the 80s, homophobia, OPEC, Vietnam, the Cold War, unbelievable pollution, smoking literally fuckin everywhere, etc etc. Going back isn't an option.

It's hard, but we can do hard things. I actually think we're in a really exciting time. We've never had more resources, we've never been better at making decisions (honestly just read about Robert McNamara), and we're newly unburdened by a prevailing orthodoxy. If we do it right, the next era could be really incredible.

Cool looking forward to reams of AI slop in a language I can't possible know. A reasonable way to live your life is to do the opposite of whatever Thoughtworks says you should do.

Fine, let em have little video recorders then

The power balance is tilted heavily towards employers though, no? If I quit you still have things like income and health care. If you fire me, not so much.

I think this is actually the norm.

My hottest policy take is that all physicians have to watch House all the way through.

No, I/P is a huge purity test for the DSA. They explicitly do not endorse people, for example AOC, for having any nuance in their position whatsoever. They are looney tunes.

https://dsaemerge.org/2026/04/05/no-iron-dome-no-exceptions-...

https://www.dsausa.org/statements/status-of-dsa-national-end...

Yeah, exactly.

To be clear, I'm not advocating for loli hentai or CSAM. I'm closer to the ban Internet porn side of things. But it's not because I think it turns you into a degenerate (I'm on a big "please interact directly with more humans" kick lately). While it's true if you're sexually inexperienced you get a weird idea of sex from porn, it's not as though you watch a lot of, say, stepsister porn and dot dot dot, a few months later you're propositioning your own sister. These are the same wrong arguments people made about video games, horror movies, whatever. It's moral panic, pure and simple. And anyone who doesn't think that should explain why it's illegal to have CSAM but not snuff, why it's illegal to generate CSAM but not pictures of people being murdered, etc.

This would make cartel execution videos more abhorrent?

I don't really wanna get into a hierarchy of abuse/violence; let's just say they're both bad. But watching a video of a beheading isn't beheading someone. We may have a moral objection to someone deriving pleasure from the watching of it, and we may even worry that without regulation we'd create a market for this abhorrent act, but it's still not the same thing.

Why would watching or possession of evidence of abuse against a person be an offense against a person?

I didn't say this. I said CSAM is evidence of abuse against a person in a way anime can't be.

I think it's perceived as abhorrent mostly because it's an evidence of the person watching being sexually interested in kids.

Yup, it's moral outrage. I'm not saying that's good or bad--personally I think there isn't enough moral outrage these day. All I'm arguing is the way we treat people who have CSAM is pretty unhinged, and it's even more absurd when we talk about AI-generated CSAM and like, loli hentai or whatever.

Privacy is a requirement for freedom. You would act differently if everyone/the government knew everything you were thinking/saying/doing. The watchers would also act differently (think about all the accusations of incompetence when people post violent things on Facebook and then go on shooting sprees).

Which is a moral affront, not an offense against a person. The whole reason CSAM is abhorrent is its evidence of abuse. A cartoon is evidence of nothing. We should treat them differently.

Frankly, I don't think anyone cares that these images are generated using CSAM or non-consensual imagery. If I generate a perfectly ordinary picture of a person and the model somehow used CSAM or non-consensual imagery to do it, no one would care. Conversely, if we could prove a model generated gross ass images using nothing gross ass, everyone currently objecting would still object. So, let's just say what we mean here: we have a moral problem with people generating, storing, and sharing these images.

Car manufacturers don't want this because it gives Apple/Google leverage over their products. The alternative is some kind of open standard, but that undercuts the ability of manufacturers to hard fuck you on trim options.

One of the more irritating parts of late stage capitalism is the complete inability of its scions (CEOs, etc.) to say any of this. The gaslighting is insufferable.

Understanding was always the bottleneck.

It drives me bananas when progressives defend corporations, particularly around things insurance (everyone knows insurance companies screw their customers as a matter of policy) or price fixing foods during and immediately after COVID. All the incentives are to do this, it happens all the time, and governments don't really want to put the screws to companies for doing it because that hurts the stock market.

You should have to have a license to be in the c-suite of a company over 200 people, and if you do shit like this, we revoke your license. The incentive structure is built on the hope that when threatened with making merely minimum to median wage, you'll suddenly care about the cost of groceries.

I love my daughter, but I’ve never been a little girl; it doesn’t bring anything back for me. It’s like Mudjekeewis says - you’ve got to have a son to see your youth rise up before you.

This feels pretty unnecessary.

The online parenting community is large, opinionated, and vicious.

Extremely super true, and it's even worse for moms. You also have to add "predatory" in the sense that people will sell you all kinds of snake oil to solve your problems that will at best not work and at worst mess your kid and/or you up.

The motivations underlying 2A are:

Protection

Participation in military / national defense

Resistance of tyrannical government

Hunting

Depending upon whether or not you think the Constitution is a living document, a modern reading of 2A could reasonably include things like explosives, drones, radar, etc., but maybe exclude things like nukes, fighter jets, biochemical weapons, other purely offensive things. I'm very pro-gun regulation, but I think this would be a fine reading as long as we're doing the same thing across the Constitution, i.e. substantive due process.

But while conservatives love modern readings of 2A, they deny modern readings of anything else. So they have to find some way to fit their desired outcomes into originals/textualism, leading to absurd dilemmas like "either the founders meant muskets or they meant nukes", or tortured standards like scanning all firearm or self-defense laws in effect around the late 18th century to discern intent, which predictably do not emerge from consistent foundational principles because their authorship is scattered across space and time and thus really are no help... unless of course you cherry pick shamelessly.

No. While originalists and textualists purport to refuse to extend any principle into the modern day ("no right to privacy in 3A, 4A, etc"), one they do is that 2A doesn't merely apply to arms of the day, but also to modern arms. It's... pretty blatant.

That it's easy to build things when you oppress and enslave people.

I don't know, Lua is way more successful than Ruby, Dart, Swift, Perl, Kotlin, etc. Something is going right (it's very good for scripting engines).

Lab leak is total bullshit. The one person who had symptoms took a train to the market, and the ride's like 45 minutes. What did she do, hold her breath the whole time until she got to the market, then cough on everyone she saw? Viral load analyses of the drains spiked under the bat cage. Did the woman, IDK just take a huge shit in that drain? It's way in the back of the market by the way, so you'd think a drain closer to the entrances would be the hot spot.

The evidence for lab leak begins and ends with "there's a virology lab in the region". Gotta do better than that.

One of the interesting things about Lua is because they don't really maintain compatibility between major versions, there isn't a huge ecosystem, and as a result there's less friction against making your own, slightly incompatible version. When you add on the simplicity of implementing the language, it's created a really diverse set of lua-alikes. Weird (and cool) for a language to have a diverse ecosystem of implementations, but not necessarily libraries.

The copyright office itself also says this when it talks about determining authorship.

I am never even once hearing intellectual property or copyright claims from Anthropic, whose product depends entirely on having consumed all human output ever made regardless of those rights.