> Car debt is terrible. But many people can’t get car debt, because many banks won’t write a $3,000 car loan, much less to someone with unsteady income. The bottom quintile often can’t even reach the $1,000 car - it’s hard to get out of the cycle where you can’t get a job because you don’t have a car. When people wave off Waymo or bus lanes or bike lanes because car culture is fine, they mean it works fine for them.
Banger post. My first car was a $1200 Geo Prism that I bought through the mechanic shop at the gas station up the street. It was older than my (younger) brother; got a good 5 years out of it before the transmission literally fell apart on Rt 2.
My grandfather bought me a $200 Ford Taurus at a car auction. I hated that car. It was this ugly green color and handled like a boat. But I drove it for many years, mostly to part-time jobs waiting tables or working at a butcher shop when I was in college. It finally met its end one winter in Worcester, Mass when the road salt ate through the flange that held the exhaust pipe on. This happened while idling in the parking lot in front of a Jo-Ann Fabrics store. If you’ve never experienced an unmuffled V6, think twin engine turboprop loud. A kind employee at Jo-Ann came up with a clever solution to hold the exhaust pipe against the chassis using baling wire so that I could at least drive it without the exhaust pipe dragging on the ground. I continued to drive that car, until my neighbors complained that the “tank noise” was waking them up in the morning. I replaced that car with another car that was even worse, but that’s another story.
I am comfortably middle class now. I still buy budget cars because I can’t get past how outrageously expensive most cars are now. But even now I still think, every time I sit in my new (well, three year old) car: wow, this is awesome. I don’t know if my friends think those same thoughts, because I would be honestly a little embarrassed to explain why I feel that way.
I think that anyone who has not experienced the slow churn of driving a car with the engine light on for years, some duct tape holding something in it, is missing nothing. But I still feel like a child with a base modern family car. It is the hint of time travel we are allowed.
It's not fun or bragging-worthy or anything. The "haha the driver door doesn't open" bits are very different from getting stranded in random places. The only good experience is if you manage to fix it for cheap.
It was cheap, but on this car, it was not easy. I tried. I would have no trouble replacing it now, but then, I had no money, no tools, and I could not afford to bring it to a mechanic.
I have one piece of junk and one newer (ie 2012) car, so I get to experience the difference every week. Except, the driver side window doesn't work in either car. Wish power windows weren't invented.
This made me laugh nostalgically. A Geo Prism was one of my first "$1000 cars" as a kid. Fun little car, but once it hit 40 it started shaking like a tree. I always joked that if I drove it on the highway it would shake itself apart around me and I would end up with suspended in air with only a seat and a steering wheel left, looney tunes style, before plummeting to the ground.
I sold it off for a different $1000 car before it happened, but sounds like that's precisely what happened to you!
It’s wild to me that the US tries to resolve labor issues with immigration control rather than just having good unions. Back in the 80s we decided we wanted free commerce at the expense of all else. Well guess what, that means you’re competing with the global labor pool. This shit is like trying to have your cake and eat it to; we want only the upside of the global neoliberal order, and try our damndest to outsource all its downsides.
I think there's a very interesting history to be written/read about the emergence of the persona of 'founder'. I'm seeing it all over the place these days (alt. 'entrepeneur' or 'changemaker', the second of which really grates on me). It feels (to me) very much like 'small business owner' or 'the kids' or something: an idea or prototype of a person who is to be managed and possibly even produced by society. Not saying this in a critical way, I think every era has it's various typologies of 'person in society', but this one seems kind of unique in some interesting ways; very much a product of the last ~50 years of US history.
edit: I'll say that in my personal opinion the critical thing a founder needs is a strong internal locus of control. I feel that this is also an important product of a liberal arts education, so I see some natural alignment. I worry about the degree to which the 'total university' strips 18-22 year olds of important rites of development: learning to feed for yourself, find housing, meet friends organically (as opposed to because you share a dorm), etc. I could see a more pared down, "education-mostly" school in a large city being an interesting avenue for new models of university.
But in the case of CPU architecture there are SOME people who understand how things work 100%, and they've built and vetted abstractions/mental models that enable other engineers and scientists to have that kind of mixed shallow/deep understanding in a way that works. On the side of LLMs we're still lacking an expertise which could flawlessly explain how these things operate; the abstractions that we're using are instead derived inductively and are totally unvetted.
You are right that the field doesn’t have a theoretically sound explanation for the architectural choices aside from “A works better than B”. However, I would argue this is an ideal opportunity for the “gentleman scientist” or eager 17 year old.
Basically every part of the original transformer was replaced with something more efficient or better:
LayerNorm -> RMSNorm
Sinusoidal position encoding -> RoPE
MHA -> GQA
ReLU -> GELU
What this means is that there is ample opportunity to improve on what we’ve done thus far.
The original RoPE paper is a great example of the authors not actually understanding how their own discovery was functioning fully which is a fun example of what I’m talking about I guess…
Yes you might progress the field, but will you really understand why? You can make up an explanation and anthropomorphise it with a few contrived diagrams and everyone will cheer!
Oh yeah I absolutely agree that there's lots of greenfield for relative newcomers to find some big improvements! I'm just saying that reasoning about 'best practices' in terms of what works for something like a CPU is missing some important and relevant differences.
Developing an intuition about high dimensional systems is pretty different from understanding the character of some specific point on a 1e9+ dimensional manifold of parameters, in my professional opinion (setting aside all the degrees of freedom that come from the structure of the thing). Sure one can understand generic principles like the curse of dimensionality, but truly groking how an LLM works is basically an open problem as far as I'm aware. I'm not saying there's no benefit for amateurs to study how LLMs work, but let's be realistic about how far mere intuition can truly take anyone in this space.
Back when I read "Cybernetic Revolutionaries" I followed it up with "People's Republic of Walmart" and came away with a bit of a different conclusion to your point 1. IMO lots of US-based hyperscalars are managing what are effectively command economies, and it's allowed them to ex. scale their infrastructure so aggressively these last ~2 years.
Right. WalMart and Amazon are closer to command economies than Gosplan ever was. They're not choking on the scale of what they do; they're thriving on it. Gosplan, the USSR's central planning system, ran on monthly reporting and annual plans. WalMart at one point ran on daily reporting and weekly plans. It's probably even faster now.
The problem of getting accurate data started to be solved when bar codes and RFID tags came in. It's possible for a factory to fake "we made 431 washing machines today." It's hard to fake "we made 431 washing machines today, their serial numbers were scanned as they left the factory, scanned again as they arrived at the various distribution centers, and scanned again when they were shipped to a customer, and scanned again when the customer received them." It's not airtight, but faking it requires a sizable fakery operation which tends to be detected eventually. Much real world activity is driven by all that "where's my stuff" data, and if the data is way off, people notice.
For large classes of products and services today, there is no real price competition. There just aren't enough players to make a market. The magic number seems to be four, from an EU study. Less than four major players in direct competition, and prices don't go down.
FMCGs work very much like command economies, including specialized tooling to shift production between customers when orders fall through.
Unlike USSR they simply do it on shorter timescales, too, and with better customer feedback. Something Cybersyn actually targeted, as it's closer to Toyota Production System (as much done at lowest level in small decision loops) than centrally managed GOSPLAN or large american FMCG giants
Yeah, the economy is basically a bunch of oligarchical/dictatorial command economies, or what we like to call companies. It's funny, a lot of leftist people use them as examples of how a larger command economy could work (usually the reaction a lot of leftists I know had to People's Republic of Walmart), whereas it makes me wonder what more democratic, free market companies would be like.
> it makes me wonder what more democratic, free market companies would be like.
I was wondering the same after PRW lol. Recently I had the chance to ask a friend of mine who worked in Google Infra (years ago) if they had ever tried to 'liberalize' their infrastructure planning; something like having internal business units bid on space, power, and compute instead of distributing it top-down. He said that yes, this was actually something like their original operating model, and the outcome was that YouTube basically bought up every piece of hardware in the company and then leased them out to other verticals at a profit.
That's pretty interesting. I always found the idea of democratizing a company pretty easy to understand (can easily just mirror civilian governmental structure in some sense), but creating an internal free market system harder to wrap my head around so pretty cool to see an example of it.
> whereas it makes me wonder what more democratic, free market companies would be like.
There are a few small examples that come to mind (co-ops and communes mostly), but I can’t think of any that have been successful at scale. Running an internal market would be almost pure overhead, so if it worked it would really have to be efficient to be worthwhile.
I understand that line of thinking and honestly think it's possible you're correct. On the other hand, this is exactly historical rhetoric that was used against republicanism/democracy, with people saying the only successful examples were city states and that it was impossible to scale. Maybe we just need a corporate "America" that figures out the correct governing structure. If anything I think it's possible it might only be possible with extremely large firms while small firms stay autocratic to stay dynamic.
Internal pricing always turns out to be problematic because the various divisions have neither the incentive nor the leverage to negotiate prices in the way that "real" customers do.
In addition to your Samsung example, I know of a major tech company whose services division often buys hardware from competitors because they get a 40% discount and priority deliveries when stock runs low. If they were to buy the in-house alternative, they'd be paying list price and be at the back of the queue for deliveries.
There's also huge scope for "funny money" Hollywood accounting-style practices, where internal prices become entirely disconnected from supply, demand, or any sort of underlying value - which actually results in the organisation being even more dominated by the whims of senior leadership than they otherwise would be.
(These problems are very similar to those experienced in country-level command economies too, of course)
I imagine it might look something like a company-internal gig economy; someone puts out a request for a code review and a bunch of fellow employees bid on doing the job, and the requester provides a tip for good service at the end, etc.
Well, at the extreme end of free market, Sears CEO Eddie Lampert drew from Objectivism and had his company's divisions internally compete with one another:
I'd honestly recommend checking out the People's Republic of Walmart; it's a super short read and anything I say would be basically a warmed up rehash of whatever I remember from it.
Basic point is just that operations research/supply chain management was heavily developed by the USSR to support the central economy and now it's used by every major retailer (among others) to manage supply chains that are massively larger than anything the USSR ever got close to.
From the article:
> We said criticality in 2026, electricity production in 2027, and power to the warfighter in 2028.
Are there any other examples of land-based militaries using nuclear power? Seems kind of like since they can't talk about the energy transition or w/e this has to be a military thing instead.
That's originally what the SL-1 reactor was meant for. It was also developed at Idaho Falls.
Instead of parking diesel generators up at Arctic radar stations (and other remote locations), we'd use a nuclear reactor, which is a logical choice.
The problems were:
1) one control rod could send the entire reactor prompt critical
2) rods could get stuck, requiring that they be "exercised" very carefully as to prevent them from making #1 happen
3) this maintenance was being done by a man in his early 20s who had been sleep deprived and received a phone call before his shift from his wife asking for a divorce
The result was the only immediately-fatal reactor accident in American history.
Glad to see that we're getting over this and moving forward with the concept. Only took us 60+ years...
"Stalin is proud to support the rebirth of Soviet Russia’s nuclear industry and ensuring soviets have access to affordable, reliable and secure energy for generations to come". Always mention the fearless leader.
Banger post. My first car was a $1200 Geo Prism that I bought through the mechanic shop at the gas station up the street. It was older than my (younger) brother; got a good 5 years out of it before the transmission literally fell apart on Rt 2.
reply