I think another source of variation on the results is the actual nutritional intake of the subjects.

The vegan/vegetarian is a big spectrum. A vegetarian eating plenty of dairy products and eggs should in theory be much better creatine-equipped than a poor vegan student feeding on noodles and ketchup pastas. A lot of pseudo-meat for vegans are full of processed shit.

Confusion will be even higher if you include fish in the vegetarian diet, which is odd but some people do that.

Could the cognitive effect works in the case of creatine deficiency (vegan) and not in the case a vegetarian (low creatine but not deficient) ?

It's OK to eat fish cause they don't have any feelings

- Something in the way, Nirvana

Could it be that the startups that embed LLMs in a product will prefer Openweights for a more stable economic model, while SWE who use LLMs as a tool prefer SOTA to produce code?

Typos are frequent (including in this blog post), and have been for a while. Bret is not an AI bot.

I'd call him a little bit sloppy, in need of proofreading. Certainly not an overuser of AI. He writes at least one of these monster posts a week on top of (IIRC) teaching in college, so it's understandable if he's in a rush.

And yeah, it's not the best, but it's really not worth discounting his writing more than he himself already does at the end. Lots of smart people have imperfect language skills.

I would describe it as entirely normal. My experience working in a research organization where the majority of my colleagues hold Phds is that education level has a strong inverse correlation with ability/willingness to care about such mundane chores as spelling, grammar and arithmetic.

A blogger.

I stopped reading when the author classified people in binary categories. Absolute proof of their own stupidity.

To be fair they aren’t binaries, they are a spectrum.

I think it’s fine to classify things like this. It doesn’t preclude the existence of other dimensions. It limits the analysis and makes it easier to understand because the point gets lost when you classify things on an n-dimensional spectrum.

Because when one exec ask: "Why is that?" the room goes silent.

Nah it just got promoted as the database for PowerBI. /s

> I think the argument that automated outputs are inherently and universally worse on objective quality measures

Just remove universally and the argument stands. The point here is to focus on the curve (it will depress the overall quality of products) not the outliers (I really like my Patagonia jacket).

I’m unconvinced: machine made fabrics go into my nicest shirts, at a price I can afford several.

How would quality be improved by making the fabric by hand? — how would I benefit from a less standard, less fine, more expensive weave of those textiles?

Machines enable cheaper goods; but the preference for buying them is down to individuals — eg, I think it’s great I can also get cheap, relatively disposable tshirts. According to you that “depresses the overall quality”; but I’m struggling to understand how the facts that a) I have better and cheaper dress shirts and b) I have cheap tshirts I don’t care about damaging when cleaning, sweating during exercise, etc together make my life worse.

Even if by some obscure metric the existence of the tshirts and my ownership of them “depresses” quality.

I don’t think it does stand. During the Industrial Revolution, automated production replaced craftsmen and often produced a better product.

Specifically, interchangeable parts were a product of industrial automation. Craftsmen were not able to build parts to the same level of precision as machines, so true interchangeability only came when the craftsmen were replaced.

As a percentage of all outputs, perhaps - But increasing to 20% high quality of 10000 vs 50% high quality of 500 would have meant an increase in absolute terms

There is more to life than Pepsi or Coca-Cola.

If it is relevant, I tend to take very good care to understand the request. I ask follow-up questions, take notes, etc. Not only it is useful (understanding the problem is key after all!) but it will show you care, and that your "I don't know" is not synonymous of "I don't care" or "Just open a fucking ticket".

Trust by Francis Fukuyama explores this relation if you want more meat than comments on a board.