Largely agree with you otherwise, not sure why you read my comment as mental gymnastics to justify LLMs. I don't think that they have an internal emotional state that can feel rushed, panicked, so on. They do — superficially — "respond" to such cues in language, which is why they can be "threatened" [0] and "flattered" [1]. But without an internal theory of mind, LLMs do this sycophantically without any internal model of the world (hence the quotes above, to avoid fully anthropomorphizing their behavior).

The only parallel I'm drawing is that both humans and LLMs can be coerced into unintended behavior via language ("social engineering" and "prompt jailbreaking" respectively), and those attacks are more effective if an attacker is allowed to control more "context", even if the underlying mechanism of why those attacks work is completely different.

[0] https://arxiv.org/pdf/2507.21133

[1] https://arstechnica.com/science/2025/09/these-psychological-...

Social engineering, like prompt injection, is a context attack — easy to spot if you're ready for it, but harder in different circumstances (rushed, panicked, tired, having a bad day, etc.).

Troy Hunt (security consultant, creator of HaveIBeenPwned) and Cory Doctorow have both been successfully phished [0][1]. They're both tech- and security-savvy people who "should have known better" but it happened to them anyway. But maybe you're different... you'd never fall for an online scam, right? [2]

[0] https://www.troyhunt.com/a-sneaky-phish-just-grabbed-my-mail...

[1] https://doctorow.medium.com/https-pluralistic-net-2025-04-05...

[2] https://news.harvard.edu/gazette/story/2024/09/youd-never-fa...

[0] https://news.ycombinator.com/item?id=48037128#48038639

[1] https://en.wikipedia.org/wiki/Peter_principle

Ideally AI would minimize excessive documentation. "Core knowledge" (first principles, human intent, tribal knowledge, data illegible to AI systems) would be documented by humans, while AI would be used to derive everything downstream (e.g. weekly progress updates, changelogs). But the temptation to use AI to pad that core knowledge is too pervasive, like all the meaningless LLM-generated fluff all too common in emails these days.

[0] https://en.wikipedia.org/wiki/Wirth%27s_law

Not quite the same, but the New York Jets (one of the few NFL teams that can match the dysfunction of the Browns — they have the longest active playoff drought in big 4 North American sports) passed on a few successful players because the owner, Woody Johnson, reportedly didn't like their Madden (video game) ratings [0]:

> A few weeks later, Douglas and his Broncos counterpart, George Paton, were deep in negotiations for a trade that would have sent Jeudy to the Jets and given future Hall of Fame quarterback Aaron Rodgers another potential playmaker. The Broncos felt a deal was near. Then, abruptly, it all fell apart. In Denver’s executive offices, they couldn’t believe the reason why.

> Douglas told the Broncos that Johnson didn’t want to make the trade because the owner felt Jeudy’s player rating in “Madden NFL,” the popular video game, wasn’t high enough, according to multiple league sources. The Broncos ultimately traded the receiver to the Cleveland Browns. Last Sunday, Jeudy crossed the 1,000-yard receiving mark for the first time in his career.

...

> Johnson’s reference to Jeudy’s “Madden” rating was, to some in the Jets’ organization, a sign of Brick and Jack’s influence. Another example came when Johnson pushed back on signing free-agent guard John Simpson due to a lackluster “awareness” rating in Madden. The Jets signed Simpson anyway, and he has had a solid season: Pro Football Focus currently has him graded as the eighth-best guard in the NFL.

[0] https://www.nytimes.com/athletic/6005172/2024/12/19/woody-jo...

[0] https://news.ycombinator.com/item?id=42505454

[0] https://www.bellingcat.com/resources/2025/08/14/llms-vs-geol...

[0] https://nob.cs.ucdavis.edu/classes/ecs153-2019-04/readings/c...

[0] https://aviation.stackexchange.com/questions/16434/why-are-r...

[0] https://www.navalgazing.net/Carrier-Doom-Part-1

[1] https://www.goonhammer.com/star-wars-armada-naval-academy-wa...

Easy fix: extend the proposal in RFC 3514 [0] to cover prompt injection, and then disallow command execution when the evil bit is 1.

[0] https://www.rfc-editor.org/rfc/rfc3514

I think this algorithm is more efficient for generating maps with only local (adjacency) constraints, but setting this up as an integer linear program and plugging it into a constraint solver is more generalizable (say, if you wanted to enforce a constraint that rivers had to flow across the whole map and could not loop).

But I agree "wave function collapse" is not really the best name, for two reasons:

- the original repository mentions "it doesn't do the actual quantum mechanics, but it was inspired by QM", but it implies something QM-related.

- as an ORIE major in college that loved optimization, I think constraint satisfaction problems are really cool and actually somewhat approachable! So calling the heuristic something else like "wave function collapse" might limit people from finding previous work and known improvements (e.g. forward checking).

[0] https://www.cs.cornell.edu/courses/cs4700/2011fa/lectures/05...

[1] https://github.com/mxgmn/WaveFunctionCollapse

[0] https://catlikecoding.com/unity/tutorials/hex-map/

[0] https://web.archive.org/web/20170218040331/http://www.dod.mi...

[0] https://acoup.blog/2021/12/10/collections-fortification-part...

> History is hard to know, because of all the hired bullshit, but even without being sure of “history” it seems entirely reasonable to think that every now and then the energy of a whole generation comes to a head in a long fine flash, for reasons that nobody really understands at the time—and which never explain, in retrospect, what actually happened.

> My central memory of that time seems to hang on one or five or maybe forty nights—or very early mornings—when I left the Fillmore half-crazy and, instead of going home, aimed the big 650 Lightning across the Bay Bridge at a hundred miles an hour wearing L. L. Bean shorts and a Butte sheepherder's jacket . . . booming through the Treasure Island tunnel at the lights of Oakland and Berkeley and Richmond, not quite sure which turn-off to take when I got to the other end (always stalling at the toll-gate, too twisted to find neutral while I fumbled for change) . . . but being absolutely certain that no matter which way I went I would come to a place where people were just as high and wild as I was: No doubt at all about that. . . .

> There was madness in any direction, at any hour. If not across the Bay, then up the Golden Gate or down 101 to Los Altos or La Honda. . . . You could strike sparks anywhere. There was a fantastic universal sense that whatever we were doing was right, that we were winning. . . .

> And that, I think, was the handle—that sense of inevitable victory over the forces of Old and Evil. Not in any mean or military sense; we didn’t need that. Our energy would simply prevail. There was no point in fighting—on our side or theirs. We had all the momentum; we were riding the crest of a high and beautiful wave. . . .

> So now, less than five years later, you can go up on a steep hill in Las Vegas and look West, and with the right kind of eyes you can almost see the high-water mark—that place where the wave finally broke and rolled back.

Hunter S. Thompson, Fear and Loathing in Las Vegas (1971)

Comments URL: https://news.ycombinator.com/item?id=45533380

Points: 5

# Comments: 0

[0] https://www.zeroto60times.com/vehicle-make/toyota-0-60-mph-t...

Comments URL: https://news.ycombinator.com/item?id=45528003

Points: 1

# Comments: 1