What exactly *are* you thinking here? That Microsoft expected and wanted Twitter trolls to turn Tay into a Nazi? That Google wanted a black George Washington? That people just expected LLMs to realize they're being tested on alignment and give different answers when they think they're being quizzed and by who? Do you actually believe what you're writing here?

I was involved in early RLHF (was writing a plugin for AUTOMATIC to let users rate responses to create an aggregate dataset to use for open-source post training). Worked for months on it. Never once occurred to me that it might make models obsequeous. I knew other people who were involved on RLHF. Not a single one ever suggested that it might. It was not obvious in foresight. When you start with "Models behave like X", you expect, in the future, them to behave like "Models behave like X, but just smarter", unless you deliberately try to change the behavior.

You realize that open source and the research community exists as well, correct?

I'll repeat: nobody was expecting this.

There is no grand overarching plan led by a shadowy cabal who has everything plotted out decades in advance with a high level of knowledge as to how everything will play out. Everyone is stumbling through the dark here. Visibility is only vaguely one step ahead.

When Word2Vec was written, the goal was text compression. The crazy properties of latent spaces were an entirely unexpected property.

When Transformers came out, it was intended to be translation software. That's it. But there were little hints at the end of the paper where they tried it on other tasks that suggested, hey, maybe this could be used for a lot more than just translation.

When GPT-2 came out, it was mostly an academic exercise, not really "useful" in most regards. But people, playing with it, started realizing that it could "almost" - not very reliably, low quality, etc - do a lot of different tasks. And it actually kinda could do some useful tasks, like summarization.

When ChatGPT came out, there were some thoughts about using it for simple tasks, but the fact that it could actually write, say, trivial bash scripts or subroutines with reasonable reliability led to a reaction of, whoa, maybe if we improve this, we can really open up a role for programming.

When the first agentic harnesses came out, thoughts of "vibe coding" were pretty far away (the term was only even coined in February 2025!). You'd ask for changes, and then review the diff; they'd do a good job with small projects but struggle more and more as codebase size grew and really needed to be babysit (honestly, it was kind of a hair-pulling experience, even though it did save a lot of time with certain things). But it was visible with each new release that the amount of babysitting you had to do got less and less, and suddenly we could see a path to where anyone can just type in something and a program comes out.

All we can do is forsee one step ahead, through the haze. We've never done anything like this before. There are no great oracles out there who can see the future here. Sorry. Everyone is half blind.

Yes, that's exactly the point: there ISN'T anything.

You're claiming that there IS. So cite the thing that IS.

Does it matter of it's true or not? Other countries will take it as licence to deploy their own weapons, or to paint the US as the bad guy.

You know they didn't know, because you've been closely following their public statements and they didn't mention it?

That makes we wonder... are you an idiot?

You're sorta right, he typed enough for you to respond.

I never put bumper stickers on my vehicle because I like to go out in the wood and park in remote places and I don't want to come back to my windows busted out or my car lit on fire.

Or worse.

It's always been like that.

In the city it's not much of a concern, and it's probably fine if you only go to popular spots in the woods with trailheads and parking lots.

What does it sound like when they "inform" you?

"Don't say the N-word around me."

Probably.

Trump knows what he's doing is extremely unpopular and so did a republicans. He knows that there are limits to how much he can screw over voters. His base will vote for him no matter what no matter how bad their lives get because of him that's only about 30% of the population. Voter suppression can make up another 5 to 7%, but that's as much as you can get out of it before it becomes obvious that your election system is rigged and the courts can't ignore it anymore.

Trump is hoping Iran would have their state backed actors do a terrorist attack and he could ride that into a third term like Bush road 911 into a second term.

We know this is the case because right before he attacked Iran he gutted several US anti-terrorism agencies and put a 22-year-old kid who used to bag groceries in charge of what was left.

There are limits to incompetence. At a certain point it's malice.

We also know that Bibi knew October 7th was coming and let it happen. You can probably guess why.

The real problem for Trump is Iran didn't take the bait. They have much better control of their assets and they prevented them from doing any unnecessary attacks. So now they control the straight and all the cards and all Trump can do is try to wait until after the midterm and then give them $300 billion US taxpayer dollars in reparations to open the straight again.

Some people seem surprised by the glasses and how they can be used in perverse ways. These people seem to have forgotten what the original aim of facebook was. The website itself started as a way for "bros" - who would have been called incels had the term existed then - to trade pics of women from their campus. How many of those pictures do we think were taken with consent? Even if they were taken in public they still could have been creepy. This is just the same mindset, applied to technology that didn't exist at this scale in the early 2000s.

No, DT is completely responsible here. Let me remind you of the history. Obama negotiated a nuclear deal with Iran-- the "Joint Comprehensive Plan of Action:-- in which they would agree not to continue working on isotope enrichment to develop a nuclear bomb, and they agreed to an international team of inspectors who could examine their nuclear facilities at any time, without prior notice. When DT got into office in 2018 he pulled the United States out of the Iran deal, and, guess what, the Iranians restarted their isotope enrichment program. In his second term, trying to get headlines to distract from other failures of policy, DT attacked Iran, saying that he needed to stop them from developing nuclear weapons. It is now pretty clear that there is no possibility of Iran EVER agreeing again to inspections; there is no way that we will ever get an agreement as good as the one that DT tore up. And Iran's attitude now is that they need to develop a nuclear bomb,

We should have quit while we were ahead, and let Iran go to hell in their own way. But that might have allowed Obama to take credit, and one thing DT can't stand is anybody other than him taking credit for anything.

Right. The increased price of oil is pure profit for the oil companies in the US. If you don't have to ship through the Straits of Hormuz or the Bab el-Mandeb Strait, you're selling your oil at double the price.

I think the subject line of this comment thread is backwards. Agents don't do anything, models do.

I'll say it again. The agent does nothing whatsoever without the model. And that goes for the API and the watchdog processes.

It's not yet well known.

Because I was closely following the field at the time?

Find an example of anyone meaningful in the field expecting these things beforehand.

Don't worry, I'll wait.