They don't, they don't have agency or intelligence.

They use more expert language when the user includes more expert terminology - because that's the nature of how they use context to generate answers.

You use more expert terms, so they draw on training data from more expert source - sources that contain more of that same language.

There is no thought or agency about it. They're not "rewarding" anyone. They're mindlessly reacting based on probability.

Historically, automation are all most efficient when used by experts.

However, the content in this blog is exceptionally jarring promotion of so called "prompt engineering".

If there are simple univeral principles that can be understood by high schooler can magically make LLM more useful, you can bet it is already built into the harness, or even model.

People cannot prompt like Tao because Tao is incredibly good at mathematics, not because Tao is good at prompting. It is very likely with enough time and energy Tao can find the counterexample himself, most people cannot.

Very long winded way of saying if you dont know wtf you are doing you won't produce anything good with AI.

Lol, no

If you have a good theory of your codebase, you can push the LLM much harder than if you have no familiarity. Because you have your own sense of what a good solution might look like, you can say “no, I think it could be simpler here”, or “but don’t we already do X?”, or “can we express this problem in these familiar terms?“.

Yes. When you already know what you’re doing, AI can work as a force multiplier.