You can't really in this way. They have a parameter they control on the backend that can force how much time it thinks for

In that case the author is simply saying the LLM is a p zombie, why does he not know what he's talking about?

They are saying something similar to "LLM has no soul", depending on context it might something insightful or (in technical/scientific context) they are making fool of themselves.

That does not explain how it is possible for a simulation to experience itself

It doesn't. It has the illusion of experience.

Not for coding because it actually needs to read and write large files

Well, sort of. Imagine the case where it first scans the repo, then "intelligently" creates architecture files describing the project. The level of intelligence will create a varying quality of summary, with varying need of deep-scans on subsequent sessions. Level of intelligence will also increase comprehension of these architecture files.

Same principle applies when designing plans for complex tasks, etc. Token amount to grasp a concept is what matters.

Tbf, I have not super kept track of what is actually happening inside the "thinking" portion of recent releases. But last time I checked there still was a lot of verbosity and mistakes, that beat the actual amount of required, usable code generation by a wide margin.

Do note that that is a different model. The one we are talking about here, DeepSeekMath-V2, is indeed overcooked with math RL. It's so eager to solve math problems, that it even comes up with random ones if you prompt it with "Hello".

https://x.com/AlpinDale/status/1994324943559852326?s=20

That's a different model: https://huggingface.co/deepseek-ai/DeepSeek-V3.2-Speciale

Oh you may be correct. Are these models general purpose or fine tuned for mathematics?