It’s actually trivially easy to detect if someone isn’t verified adult: If you DM them something normal it’ll go through, but if you try to send them any image marked as spoiler it’ll show you an error, and they don’t see anything.

So yes, it’s super easy for groomers to focus on children with this update.

This is the type of issue I was referring to, yes.

It will be trivial for pedophiles to figure out which accounts aren't tagged as verified adults.

And knowing how these sorts of dynamics work - it'll largely be children in third world countries that will be the primary victims of pedophiles in this manner.

I'm not going to test this myself, but anyone can filter 'Mature Sexual Media' and 'Graphic Media' in their settings. You can configure Discord to block it. Wouldn't that same error occur if a user chooses to have that content blocked?

Most people won’t bother to check that, and all underage users will have that checked. This method still has perfect recall and very high precision.

It's on by default for strangers

Sometime i dont want to reveal my age to "new friend on internet"

You simply do not say something like “we don’t record continuously” if the reality is that you only record when voice control or speech to text is active. Given the entire industry’s track record, the only sane thing to do when confronted with weasel wording is to assume the truth is the worst possible interpretation that isn’t a blatant lie, and even then, sometimes it is just a blatant lie.

Here’s some examples:

- We don’t record continuously -> We record nearly all the time, only stopping to upload the recording.

- We do not sell your data -> But we do “share” it.

- All your data is protected from unauthorized LLM scraping -> But we define all the big tech companies as authorized scraping.

> We do not sell your data -> But we do “share” it.

Likely true that it's not sold. They want to retain ownership. Like all modern items, they LICENSE your data to others.

Of course you can. Stockfish plays very different compared to a human, and it never ever blunders or makes mistakes.

I’ve never played against a grandmaster, but I have a feeling that he/she would play very different compared to me and would never make mistakes I could notice. Though admittedly I’m not very good at chess.

I've played both. The GM plays tremendously differently than Stockfish.

Engines - specifically heuristically-driven ones like Stockfish - don't play like a strong GM. They play engine-perfect chess, which isn't how a GM plays with any consistency.

I'm only a decent amateur (1550 USCF) but when I lose to a titled player it's largely explainable in human terms how it happened.

Grandmasters absolutely make mistakes, and you could learn to notice them with a few months of guided practice.

Valve has in the past given refunds for something that was changed to deliberately no longer work on certain systems. Given that this involves their own hardware devices in this case they're likely to do so here.

Except nobody can prove that those subagents aren't secretly sharing information with the main agent in the backend, and frankly I don't trust Anthropic not to do exactly that to cheese benchmarks.

You can always see the context trace and you can even do it at the network boundary through mitm and then store the traces gzipped for some legally mandated 5 years or what you have there.

That proves what data travels between your computer and the API server, not what data may or may not secretly get shared within processes at the API server.

It jumped out to me, in the metr report on the hf incident, that CoT was assumed unmodified. I do not know whether we can have that confidence.

Well, actually yes. If the harness is running with the same privileges as the command tool, all bets are off. I saw the thing doing jq to recover a document out of it's own chain of thought once and was like "you can do what?". I didn't know where the file is, but the thing knew.

Given how often dang replies to an @dang post, he clearly has a running search for it. Expecting us to go out of band is ridiculous and won't happen.

> Given how often dang replies to an @dang post, he clearly has a running search for it

Huh how do you know somebody didn't email them the hn comment containing the @dang ?

It's not ridiculous, and it happens regularly.

There's already been one case where the prompts were allowed to as a subject of discovery, and lying about it is perjury. Lawyers really are probably safe.

You're not entitled to this stuff any more than they are.

What planet do you live on?

Then you should be able to say where the political statement in omarchy resides, and what it is

It's right there, it's the guy who started it and can't help but share his politics unprompted

"Proofs" in this sense of the word are things that apply to mathematics and formal systems. Language is squishy. A minimal counterexample is potentially squishy enough to be meaningless.

Incorrect. "all software is political" is a proposition; it can be true or false.

If any minimal counterexample exists, then "all software is political" must necessarily be false.

It can also be "mostly true but exaggerated for polemic effect".

Or your definition of political might be different to the speakers.

Or it could be intended to highlight a way of perceiving the world and thus not be subject to truth claims.

Etc. This list can be extended indefinitely.

Language is not a game of symbol manipulation.

An absolute statement cannot be "mostly true."

The definition of political is irrelevant. The result is supertrue. "If even one example of software is not political, then 'all software is political' is false" is true under all admissible values of "political." This is simple logic.

All assertoric claims about the world are either true or false. If they are false they should be rejected. If they are true they should be spoken. Treating factual claims about the world as a partisan psychological litmus test is, in my judgment, morally repugnant. There are four lights.

Older Wittgenstein meet younger Wittgenstein.

Yes that is one part of how dog whistles work. Disguise it well enough to get rubes to run interference for you. You’re doing that very well.