I also thought that was weird. Then I learned it gets better. If you click through to the BBC article that was apparently their main source, the quote is this:
> Alternatively, as Prof Barber explained, it can be compared to a single string of spaghetti holding up 3,000 half-kilogram bags of sugar.
So the professor used an item that was familiar to his English audience (1500 kg=3307 lbs), then the Smithsonian writer tried to be helpful in converting the units, but switched to an item far less familiar to an American.
I don't think I've ever bought a 1lb bag of sugar here, while a 500g bag is a little small but normal in the UK.
Except, a gallon is a gallon no matter which gas station I'm at. Also I know my car's gas mileage, and it doesn't change when I visit a Shell station instead of a Chevron. The composition of the gas is regulated, as are the pumps that dispense it. There are inspectors from the state whose job it is to ensure that when I buy a gallon, I really get a gallon.
Tokenizers aren't standardized to anywhere near that level. A "token" from one isn't the same as a token from another.
It's not that bizarre. Not everyone is trying to optimize for maximum profit. Some creators or companies want to build a community by increasing the consumer surplus received by their buyers. They are willing to trade profit for that. Scalpers slip into the middle, take the surplus for themselves, and prevent the community building or other social goods that the seller is trying to create.
"SUV" is too broad a category. A RAV4 is an SUV. It's similarly sized to most of the SUVs I've seen in Europe. And a pedestrian getting hit by one would have a similar experience to getting hit by a sedan. It's nothing like the big Rams, GMCs or F250s with the high front grilles that are becoming more popular while also being far deadlier to pedestrians.
I agree with the "steal an ancient roman baby" premise. The "roman citizen" example is not as strong. Cognitive ability is not just genetics. The grown-up roman would be missing a lot of advantages during their upbringing that weren't available back then. Also, limiting it to just "citizens" means limiting it to their upper class.
Compared to Roman times, we've had pretty big advances in nutrition, healthcare, education, and widespread middle class wealth. It's not unreasonable to infer that these would have an impact on cognitive ability similar to the effect they've had on life expectancy.
That being said, there's definitely a present-ist bias, as the McSweeney's article does a good job mocking. I do believe their best thinkers were as good as our best thinkers.
Good piece, but I think there's a missing angle to it. He cites a study showing how often people say they "use AI", and a little over 50% use it less than once per week.
If we're just talking about AI chat interfaces, sure. But I think the way that AI usage is going to grow isn't mostly by getting more chat engagement. It's about baking AI features into software that people already use.
For example, suppose you asked the same people "How often do you search on Google?" I am willing to bet the numbers go up a lot. And all of those people are "using AI" in a very real sense, they just don't think about it when it's baked in.
If you've ever watched an episode of "How It's Made" and seen how incredibly customized these machines are, it won't be surprising that the people who build them are proud of their work.
I don't think handmade shoes are the right comparison, unless you're talking about software that's meant to be used by just one person.
The more apt comparison might be to the engineers building and maintaining the manufacturing equipment in the shoe factory. It's a degree removed from the consumer, but there's definitely still craftsmanship involved there.
They tended to be solo productions, or sponsored by aristocratic patrons. Anyone suggesting that we could create movies, TV, music, or games on the scale we do today, without copyright, does not seem worth taking seriously.
The Claude Code client source was never their moat. There are plenty of other companies with equivalent tools (gemini cli, cursor cli, augment, codex, etc.) The models that it talks to are far more important.
Not to say you're wrong about commoditization. I don't think these companies will be able to raise their prices and keep them there to make enough money to keep building models like they've been doing.
You're saying no one capitalizes software development costs anymore? The rest of the internet disagrees strongly. The finance team at my employer would disagree strongly. Accountants talk to dev teams all the time about how many hours went into new development vs maintenance. It's not just a temporary IRS rule. It's GAAP.
It's the same in other industries. In your example, if labor went into building a whole new assembly line that pays itself off over several years, capex. If it just goes into building products to sell, opex.
There are licensing laws already protecting the lawyers whose names appear on motions and briefs, but not much protection for the junior lawyers who will be impacted most. Big law, like the fancy consultancies, was historically built like a pyramid, with an army of 1st-3rd year associates doing due diligence and document reviews. The bottom was cut out of that in the 2000s by offshoring and automation. AI is contributing to another wave, but not dropping off a cliff.
No, not quite. It really comes down to opex vs capex and the depreciation schedule for your investment.
Software development is typically categorized as capex, on a 3-5 year depreciation schedule. You assume the software you write today will be generating value for you that long.
If a big, expensive model training project only gives you value for a year or less, that is not like most companies.
Macs are very popular in schools today for teachers and staff. Switching to Macbook Neos for students would actually simplify their support burden. I'm not sure they'd be cost justified though.
Still if you're the lawyer on the side of the lawsuit claiming that the code is copyrightable, you really don't want that copilot attribution in the commit message muddying the waters.
> Alternatively, as Prof Barber explained, it can be compared to a single string of spaghetti holding up 3,000 half-kilogram bags of sugar.
So the professor used an item that was familiar to his English audience (1500 kg=3307 lbs), then the Smithsonian writer tried to be helpful in converting the units, but switched to an item far less familiar to an American. I don't think I've ever bought a 1lb bag of sugar here, while a 500g bag is a little small but normal in the UK.
https://www.bbc.com/news/science-environment-31500883
https://www.sainsburys.co.uk/gol-ui/product/sainsburys-white...