msteffen

Ok, I did more reading. I was actually super wrong about this, but it took me a while to get there. Oops! But very interesting!

Basically, people and firms are little black holes of unmet demand for all sorts of things. When the productivity shock happens, demand for stuff in the now-high-productivity sector(s) is now (mostly) met, freeing up money (that people and firms in the high-productivity sectors have more money to meet their own latent demand is also true, but somewhat incidental). That money goes to meeting other latent demand, some of which is for stuff in low-productivity sectors (housing, childcare, education, healthcare, construction), and that’s what drives up prices until/as people eventually move into those sectors. Even if housing specifically isn’t constrained, it still happens.

And I guess…as wages go up, some forms of automation also become economically sensible, which is what changes the balance of low/high productivity sectors. Latent demand for rope-holding is met with cones, latent demand for maids is (mostly) met with washing machines and dishwashers, latent demand for secretaries is (mostly) met with personal computers, etc.

So with AI, it sounds like people will move out of software into construction, healthcare, etc (only specialized work where the automation isn’t “good enough” stays). All the “teaching kids to code” that schools are doing now, eventually turns into people dealing with their own AI output as adults, where previously they would’ve hired a developer (like how managers who used to have a secretary now get a computer and type for themselves). Wages go up, education, healthcare, and housing get more expensive.

TIL! I’m entirely inexpert in the topic and posted as a curious layperson.

That said, some of what’s in the article only makes sense to me in the context of a system that’s at least somewhat like what I tried to describe. Maybe you can enlighten me?

> This effect is an example of cross elasticity of demand.

Is the cross-elasticity of demand here not referring specifically to the cross-elasticity of different forms of labor, which exists due to the many common inputs of different forms of labor? Eg land, healthcare, unskilled labor (as an input, in the case of skilled labor as the output—doctors and engineers in my comment)?

In the case of a quartet, if musicians charged what they did in 1700, they would quickly become homeless, and then quickly get other jobs.

So people are either leaving the profession or raising their prices, and either firms pay, high-wage-earners pay, or they get no customers and again get other jobs (like the rope-holders). Is that not the effect?

I think this is what Baumol’s cost disease is about (productivity growth on one sector drives up labor costs and ultimately prices in other sectors, such as healthcare, construction). Basically, when software or AI is super lucrative,

1. students go into new lucrative thing

2. supply of doctors goes down

3. doctor wages go up

4. hospital margins go down

5. Hospitals raise their prices to stay in business

6. Wealthy engineers are able to afford higher prices

7. Services that are transitively consumed by the high-productivity sector stabilize at a higher price. Services not transitively consumed by the high-productivity sector go out of business.

- I remember visiting a developing country about 15 years ago and seeing people getting paid to stand there and hold a rope, because I guess their time was cheaper than cones, which I think is the inverse of this.

- I also think that, in the limit, society consists of a handful of economically viable occupations and infinite cheap stuff. Nothing else is worth the prices that a seller would have to charge to afford employees.

- I also think that we’re entering a stabilization phase.

Yeah, I forget the wording but there’s a notification like “so-and-so’s AirTag is near you and they can see where you are.” My spouse put AirTags in our baby stuff and I get them whenever I’m using it.

> At this point, it seems more a discussion of morals and social values rather than a scientific "how to lose fat".

He uses language like that in the book (and I was speculating about “what he would say”).

I think the reason is that when people hear the theory “UPFs caused the obesity epidemic,” their brain sometimes goes to “well what is in UPFs that causes obesity?” His answer is “nothing.” He’s trying to emphasize that the issue isn’t some toxic ingredient, but that UPFs are a technology that are being used in a way that leads to this outcome.

(though the research in TFA might actually challenge the “nothing,” by showing that there is some kind of toxicity)

> That's the point at which it becomes highly speculative.

The “…in order to sell more product” part is scientifically speculative, I agree—the book makes the claim, but the scientific research it references does not, IIRC. The book’s author imputes it from interviews and the like.

The “overwhelm…feedback mechanisms” part is mainly substantiated by a 2019 study by Kevin Hall where, similar to TFA, they had people essentially live inside a calorimeter and eat calorie- and nutrient-matched meals that were UPF or not. They found that when given UPF, the same people ate more and faster, and gained weight at a pace that surprised the study authors (who were initially skeptical of the theory and had constructed the study to try and disprove it. The study author is interviewed and explains why he found it compelling).

Whether you find the evidence compelling is, of course, up to you :-), but I felt the book did a good job interviewing members of the obesity research community and tracing the history of the theory.

The book discusses this a bit, actually. It basically credits the UPF theory to one Brazilian researcher, Carlos Monteiro, and discusses how he developed the theory by applying epidemiological techniques to the obesity epidemic in Brazil. It gives the specific example that, in Monteiro's research, if a family bought sugar at the grocery store, it actually reduced the likelihood that they were obese because, even though sugar is not healthy, it meant they were still cooking at home.

My understanding is that he would probably make similar predictions about your examples: people who eat fried chicken and chocolate as part of a mostly home-cooked diet are less likely to be obese than people who drink diet coke as part of a diet that mostly comes out of wrappers. He would probably also say that there's nothing inherently bad about diet coke, but there is something bad about getting all your food from people trying to make money instead of from people who love you.

For anyone who hasn’t read it, I highly recommend the book “Ultraprocessed People.” It's a scientifically literate and yet quite readable overview of research into ultraprocessed foods.

FWIW there are four categories:

1 is whole foods (eg fruits, nuts)

2 is stuff that is processed but still basically an ingredient (eg flour, maple syrup)

3 is combinations of 1 and 2 (most restaurants are here)

4 is ultraprocessed food (UPF): things that were not food that have been turned into food (modified starches, artificial flavors, etc). Also, anything containing UPFs is a UPF, so most breakfast cereal (food coloring, flavors)

The basic thesis (which this research challenges, interestingly) is that UPFs aren’t inherently bad, but food scientists are shockingly good at creating highly specific and desirable flavors and textures (the book gives the example of a food company creating "ice cream" out of various modified starches and flavors so that it doesn't melt during shipping), and foods that come out of this process have generally been engineered to overwhelm our natural feedback mechanisms against overeating in order to sell more product.

I wrote a blog post making some very similar arguments earlier this week in response to some mathematicians’ takes—just before the OpenAI paper drop a few days ago (https://creating.software/essays/ai-and-math/).

I feel we’re all realizing, at the same time, that (1) intellectual activity is communal, and yet (2) those communities are sustained by recognition of individual achievement. What a way to find this out.

I claim (here and in my post) that humanity actually can’t advance without these human communities. It’s critical that people not give up learning things that AI knows better than they do, because they play an important role in these communities, and the community is critical.

My (maybe naive) hope is still that people find ways to keep doing this stuff for fun, like chess (competitions, awards, coaches, I don’t know). We still have chess competitions and the like, even though computers are far better at chess and have been for a while. The community and the recognition of individual achievement are both there. And human chess and chess engines have gotten vastly better, so computer dominance failed to kill progress in chess (today’s stockfish would’ve beaten Kasparov in 1997 running on a desktop from 1994: https://www.lesswrong.com/posts/4MLBK7iCW3vYd93Mn/a-closer-l...)

Maybe math olympiads and programming contests live on well past the end of their practical utility—I always liked them.

As social commentary, I think a lot of people in this thread are expressing interest in and engaging with this level of math who might not have pre-AI.

I bet, for people who don't understand these problems or their solutions but are close and are now interested, AI makes then considerably more accessible than they would've been previously, and behind this big visible wave of results there actually will be (or already is) a wave of improved comprehension by a lot of curious people.

I'm not at all at this level at all, but I did learn quite a bit about polynomials over fields yesterday.

My understanding was that they take issue with the word "attributable". They don't attempt to propose a mechanism by which living near Costco causes cancer (the most strongly associated of the landmarks they measured). Rather, the implication is that it doesn't. Like I think it's a lot of words arguing that "correlation doesn't prove causation" could be restated "association doesn't prove attribution."

Heterodox argument: a big part of the reason JavaScript has been so successful is that interpreted languages actually can be extremely fast—competitive with, and in many cases, faster than compiled languages, with V8 arguably the fastest interpreter in existence.

See this post by Mike Pall, author of LuaJIT (which is comparable to or faster than V8 performance-wise, despite being basically a single-developer project), which explains why: https://web.archive.org/web/20180603053407/http://article.gm.... Basically, it’s much easier to add high-quality runtime-trace-aware recompilation to a JIT interpreter, which a non-trace-aware compiled language will often not be able to beat.

I liked this a lot. I have also struggled with this, and I feel compelled to suggest my own angle: this is often caused by issues that are not (just) professional skill gaps, though closing those gaps can help resolve it.

Avoiding status updates, needing to “prove yourself,” things like these, were all products of a deeper lack of trust for me. TFA alludes to this with “imposter syndrome”. I was so used to this that it took a long time to figure out. What helped me (and I wouldn’t say I’m better, but better than I was) was a few ideas:

1. Trust is equal parts competence and communication. Trust is lost when you expected X and didn’t get it, which means half the problem is not getting it and the other half is expecting it in the first place.

2. So trust requires clear communication, and clear communication requires trust in the other direction. At a totally dysfunctional company, all of engineering doesn’t trust leadership, so they're not honest with leadership, and soon enough leadership doesn't trust engineering, if that hasn’t happened already. So lack of trust in one direction breaks down trust in the other direction.

3. If you don't feel comfortable saying something that's true (anything! “I got nothing done yesterday because I still don’t really understand the problem I’m solving,” or whatever kind of thing you’re afraid to say), it means you don't trust whoever you're talking to. This is a problem—see #2. In the context of TFA, this is the problem that leads to you skipping status updates or going off and building without prior discussion.

The solution is building your own trust, which can actually be hard to do and requires some self-knowledge (e.g. did you have a bad experience at a prior job? Are you dispositionally a nervous person? Did your boss/leadership break your trust in some way already—large or small?).

This is IMO what TFA is really about, and the strategy it suggests is a great one if a bit of positive feedback and encouragement is what helps you extend trust (which is fine! It's worth knowing that about yourself) and grunt work earns you that feedback (which it might, though you can't control what other people do—it may also help to communicate that you're looking for that, e.g. "I want to see what kind of contributions are celebrated here, and I thought these might be." A small trust exercise in the context of the larger one.)

It might be something else. Off the top of my head: will work respect my needs if a life event temporarily interferes with my work? If I'm missing context on a project, will I be given help or hostility? Am I being given a clear explanation of the expectations for the role?

(Finally, I think it's worth mentioning the version of this where your boss or leadership isn't building trust in you, and IMO it looks like you stuck in a marginal role and not given a path to larger ones, even though you're clear that you want them. Sometimes you can convince them to trust you, but you can't control what other people do. In particular, if you're building something without prior discussion, consider that a lack of trust from the company to you is now creating a lack of trust from you to the company. Maybe it'll be fixed, but no guarantees.)

> The value to mathematicians is knowing that the conjecture is probably right, and knowing the rough path the LLM has taken to it.

Ah, or is the value to mathematicians that their LLMs can build results on top of this? (In which case, did this do more than save them some tokens?) Or is the value the deep mathematical insight that this result incidentally gives a few mathematicians the confidence to develop for their own personal satisfaction (e.g. if they decide to go and prove it for themselves, and come to the same result after a lot of work)?

(IMO, the deep, scary question: what if it’s soon impossible to make anything at all that anyone who doesn’t know you personally would bother to look at or use? https://www.smbc-comics.com/comic/crack)

> Your comment is a personal attack

I don't think stschaef is a bully; in fact, if my guess that he's a researcher is correct, I think he's likely highly altruistic (I've never met him, but categorically, researchers are people who chose a difficult, low-paying job doing work of great societal value).

However (again if my guess is right), I think he could easily be in an environment where narratives about people's work being worthless, about them being stupid or unserious or otherwise beneath consideration, are common. It's a reaction to the fact that in any field, the amount of research produced is overwhelmingly more than anyone can digest. There's a lot of unstated anxiety and guilt about that, on the side of both writers (who worry no one will read their research) and readers (who feel obligated to try to read everything and are eventually, inevitably overwhelmed), which IMO is itself a product of the basically altruistic nature of most researchers.

The overwhelming reality of being a researcher is an inescapable, empirical fact, but the narratives people create around that reality, about peoples' work and its worth, are not. People are hard-wired to be sensitive to rejection, because humans are a cooperative species and social acceptance is existential to each of us, and the problem is that a lot of researchers, who are steeped in these narratives, are trapped in a self-reinforcing cycle of community attachment threat: their research sucks (or could start to suck if they ever went on vacation), and the research of most of the people who are evaluating them sucks too.

What if these stories about the worth of people and their work aren't true? Guess what: one can do amazing research and it still goes nowhere, because ultimately it's not possible to control other peoples' behavior. If one's goal in research is acceptance and respect from the community, they should consider that they're gambling their time, energy and youth on an outcome they can't control. The research community can be a fine place, but it's not special—one does not have to be a researcher, and if the experience of being a researcher sucks for them, they should quit, because living a good life is their responsibility.

I grew up around this attitude, and children are particularly sensitive to attachment threat. It's bad enough that researchers tell these stories about each other, but once these narratives and communication patterns about peoples' work and its worth leak outside the context of the research community, they run right into the reality of human attachment and the expectations of communities that aren't the research community. I'll say about my own family: I think they were good people who learned an unhealthy, judgmental attitude (towards themselves as well as others). I think they immiserated themselves (and inevitably the people around them), because perpetual attachment threat had traumatized them into false sense of obligation, and I wish they had quit.

Even if one stays, they should understand the mismatch between the narratives and culture of the research community, and they needs and expectations of most people outside of it. That was the context of my comment. I think it's fine if stschaef doesn't like TFA and doesn't find Bend novel or interesting, but my analysis of the sentences quoted reflected my organic reaction to them, and I stand behind it and my other feedback.

I find this whole post fascinating in the context of https://news.ycombinator.com/item?id=49738091 and particularly this excerpt from Gowers:

> Instead, I have a more complicated view, which I actually expressed in my essay The Two Cultures of Mathematics a quarter of a century ago, and which can be summarized by saying that there is a spectrum of attitudes in mathematics to the relationship between problem-solving and conceptual understanding. At one end of the spectrum you have mathematicians who are primarily motivated by the wish to solve problems, who see conceptual understanding as a very important means to that end. At the other you have mathematicians who are primarily motivated by the wish to attain conceptual understanding, who see problem-solving as a very important means to that end.

Before, understanding and problem-solving-ability were so interdependent that distinguishing between the two was practically very difficult and probably wouldn’t have changed anyone’s research agenda. Now, they’re not connected, and this guy just did the ultimate meta-experiment of seriously undertaking a project that is intentionally 100% problem-solving and 0% understanding to prove it (maybe 99% and 1% but pretty close. In his transcripts, he never asks ChatGPT about the math, only about its opinions of the math).

As we (as a society) sit around asking ourselves what mathematicians (and software engineers, and anyone in deep technical fields) should be doing all day, we now have this case study to show us how wide our range of options has become.

To pick on a few examples:

> It sounds quite unlikely that a random vibecoded project with a Fable-written paper has worked out all of the kinks

> If this weren't slop...

> I'm glad you're having fun vibecoding, and I like that you're interested in this area of research/engineering, but you are wildly overstating what you have here and sound sus af

These impute both his motives ("fun") and particularly his level of seriousness ("random vibecoded project" and "I like that you're interested in this area of research/engineering"—imputing passivity, as opposed to "are studying" or "are researching," which would be more appropriate given the amount of time invested).

I would actually regard this as bullying. Some feedback.

(I suspect you're an academic, either a researcher or student. I know from my own experience that bullying is endemic in many academic research environments, so if you find the negavity you're receiving "strange," I suggest finding a therapist, who may help you understand how your communication habits could be negatively affecting other people and unintentionally damaging your relationships.)

I haven’t played The Store is Closed, but to everyone saying IKEA had some obligation to shut down a parody…videogames parody real brands all the time, including their colors? Why was this different?

Some examples offhand:

- Fallout’s Nuka Cola is an obvious parody of Coca Cola (also same colors: https://en.wikipedia.org/wiki/Nuka-Cola) and Cram is an obvious parody of Spam (same colors)

- GTA is full of these: Sprunk (an obvious parody of Sprite, again same colors), Fruit (obviously Apple—same colors, same branding style, same use of iPrefix and “think” marketing), definitely others

- Saints Row “Freckle Bitches” was a fairly obvious reference to Whendy’s (“I’m all grown up”)

All of these games were bigger, not Indy games at all, and none of the parodied brands saw any obligation to sue AFAIK. Would IKEA have gone after Rockstar, given that they’re much bigger and very popular? I’m suspicious that this was a case of “small target” rather than “big problem”.

(Ok, I clicked through the article, it looks like it was just a demand letter? My impression is that demand letters are nearly meaningless. No judge has to approve them, and they cost an hour or two of lawyer time.)

> whether an autonomous running task can continue after being cancelled shouldn't be

Idk, I think this might be quite important in servers, or any application where the running task might spawn subprocesses (can those continue autonomously?) or block on/use os resources

I love this framing. I have a lot of ideas about why this is true, but the best (IMO) pat explanation I've been able to come up with is:

1. Much of constructing software is math of a sort. In math, you write proofs that say why X is true. In software, you write code that guarantees X will be true (e.g. "the backend assumed a user ID was always available, but now that we have service accounts, we need to change the access control code so that a sensible view is still returned")

2. Math is hard, and most of the work is invisible thinking. If you asked a mathematician for an estimate on when the Reimann hypothesis will be proved, they would laugh at you. Our problems are generally easier, but they can still be hard. And as in math, they're sometimes much harder than you expect (e.g. Fermat's Last Theorem. "Why was it so hard? Did you talk to Fermat? He left? But he said it would be easy!")

lol, we’re all writing the same thing. I’ll give you one example, because I particularly liked it. My therapist always talks about separating people from problems, and I never had any idea wtf he was talking about. And then he gave me this exercise where he was like, “let’s construct a narrative where this person’s eccentric personality makes them really good at something specific. Where would you expect a person like that to encounter friction and how would you manage it?”

It was sort of fun, we talked about how Arthur Schawlow (who co-invented the laser and won the Nobel prize) was banned from driving on Stanford campus and had to be chauffeured. And I finally understood what he was talking about; you can access this perspective for anybody by constructing such a narrative. What if everybody’s a misunderstood genius at something? Why not? How would you live in such a world? One insight among many.

Therapy, which (at least for me) has focused dually on adhd management strategies that can be very helpful (including, in particular, enhancing relevant social skills related to setting expectations and boundaries) and on identifying and modifying negative self-image issues and counterproductive learned behaviors (e.g. various forms of self-directed hostility that attempt to assert control over a neurophysiological issue).

It’s funny, I went to therapy a lot of times before I found someone I could stick with, and if I were to give my past self a list of what I’ve gotten from therapy, I think I would’ve been unimpressed. Nevertheless, I’ve noticed that I’m more confident and calmer, and it has definitely improved my relationships.

There’s a kind of virtuous cycle that happens where you replace a grumpy mindset where you’re always mad at yourself for not meeting other peoples’ expectations and always mad at other people for making your life hard, with a confident mindset where you understand yourself, are therefore comfortable relating to other people about your abilities and needs, and can focus on putting yourself in situations where you’ll be successful.

1. At an old apartment complex, there was a little playground in a common area, and it usually had at least one family from the complex there at any given time between about 4:00pm and 8:00pm. It was a terrible playground—small, poorly-maintained, surrounded by roads with fast cars and no fence—and people were still there all the time. I’ve been looking for an even better version of that ever since, but have never been able to find it. I wonder if part of the problem is just land use.

2. The author is surprised that so few other parents took up his pick-up/drop-off offer. I would not have taken such an offer from most of the other parents I know, because I don’t trust anybody I don’t know very well (including other parents) to watch my kids. IMHO, the author’s surprise at other parents’ reluctance doesn’t help (it makes me worried that he doesn’t appreciate how much trust he’s expecting to be handed over), even though I agree with his overall point.

I don’t think this is irrational neuroticism. I think we’ve all—guests and host in this story—lost necessary social skills and cultural cohesion, and getting them back is an important but large project.

To try and illustrate the point with an example: one of my kids has an egg allergy. A group of kids, good as they may be at looking out for each other, is not going to be tracking that. Most parents I know have never thought about what foods in their house have egg in them, or about the egg ladder and which foods are highest risk—why would they? If we were all old friends or in a close, shared community, I think this would come up. By a 7th birthday, this would’ve come up a million times and all sides would have worked through all the implications, and the other kids would know about it because they got a lecture from their parents to be careful with mayonnaise or something, and the host would know there’s an EpiPen in her bag that he might have to use on her if he’s not careful. Multiply this complexity by the number of kids at the party.

I wish we were close enough with the parents we know to understand their kids like our own. I don’t know what the author’s relationship is like with the other parents in the story, but I see in this article a bit of naivety to the reality of what the author is asking for, and I think the article itself is a symptom of the problem it complains about.

Uh, I think at the very least you have to caveat on the age of the kids. Babies need a lot of fussy obsessiveness. Kids arms aren’t long enough for them to wipe their own ass until they’re about 4. Don’t know when you get them to start doing laundry or dishes but that takes longer still.

> without much concrete evidence that it hurts long-term outcomes, assuming the child is otherwise healthy and well taken care of.

Not to pick on others’ life choices, but I don’t think that this particular assertion is true. Depends on your evidentiary bar, I suppose, but just from googling it:

* Mass General’s Fatherhood Project (https://thefatherhoodproject.org/schools/the-research/): “Findings from the rapidly growing science of early childhood and brain development show that a father’s active participation and emotional engagement with his children leads to improved social, emotional, behavioral, and academic outcomes.”

* NYTimes (https://www.nytimes.com/2026/02/03/health/fathers-child-heal...): “Research Finds Interaction With Father, Not Mother, Affects Child Health”

* https://firstfivenebraska.org/blog/6-ways-children-benefit-f... provides a catalog of evidence and reported benefits

Preach!

Bryan Cantrill (inventor of DTrace and cofounder of Oxide Computer) comes to mind as someone who seems to do both. It’s hard to know somebody’s parenting quality from the outside, but he talks about his kids a lot.

In his case, I think he didn’t start his company until his kids were older. I’m not sure what the early childhood years were like for him.

For “parenting” and “startup”, I would say that each is extremely hard to do well even when it’s the only thing on the list.

I saw an interesting article in the Atlantic[^1] somewhat recently predicting that the people who will thrive in the AI age will be ones who enjoy mental effort. That is, even when AI could do something for them, these people will choose to do it themselves if there’s something to be learned from it.

This post seems to illustrate the point perfectly to me. AI wrote the proof. It was done, Lean checked it. And presented with that, Tao’s reaction is still to want to learn something—how to solve the problem himself—and then to meticulously untangle a 90k-line machine proof (utterly disregarding that there’s no clear upside for doing so—he can’t get a paper out of this) because it’s the only way to learn that. My bet is that it was worth it.

(I also think everyone saying “it doesn’t make sense to write code anymore” is crazy. The best learning tool of all time was just invented, and you want me to not use it? What the point of any of us if not to know things?)

[^1]: https://www.theatlantic.com/ideas/2026/06/ai-open-ai-anthrop...

I know less about this than every other commenter here, but both weather patterns and market performance do seem predictable based on past behavior when modeled at the right level of abstraction. “Sunshine on Monday” does not imply “rain on Tuesday”, but “cold front moving in Monday night” does. (Likewise “stock up Monday” doesn’t imply “stock down Tuesday” but “CEO arrested for fraud on Monday” does.)

I think this is relevant to the discourse on LLMs/programming because for months, people said “they’re just regurgitating their training set,” but now I think people are seeing (I am seeing) that they do learn more abstract models of the world than that. I don’t really know how, but it’s why they can generalize from other codebases and tools and so on.