I can't claim you are wrong. But this is so non-intuitive to me that it hurts. I get why "covered ground" for text books would be easier to do. But many textbooks should be pushing the field, as well. Right?

And I fully note that the book I linked to is exceptional. It won a oscar.[1] And was basically required reading at pixar, if I recall.

[1] https://www.youtube.com/watch?v=7d9juPsv1QU

There is nothing original in textbooks? They just need to cite the relevant papers? I find it perfectly intuitive that it is harder to come up with an original argument/idea than it is to summarize one you've already heard.

Papers are very condensed and probably scrutinized more than textbooks are, via the peer review process. Papers have to have proofs for any theorems, whereas textbooks might omit them (especially the less important ones).

Also, to be fair, a textbook might rely on a dozen different papers in the span of a single page (just taking key insights or arguments for the papers).

Both of those claims seem... off to me. Specifically, this is a reductionist view of them that makes them sound easier. I do not claim they are harder, just different. Each difficult in their own way.

I would be interested in seeing the publication times for many text books. Empirically, we should see something in the difficulty in relation to the time it took to create. Right?

Edit, since I saw your point of referencing many papers in a single page. Don't papers do the same?

We were comparing the amount of time it takes to write a textbook versus the amount of time it takes to write all the papers the textbook cites (or relies on, I guess). So I don't think "hardness" and the fact that papers also cite other papers really changes this comparison.

I mean, a textbook might take years to write/edit (on and off) But it takes at least a month (on and off) to write/edit a paper.

But you don't think this is a learnable thing? Doing a paper, you already have to write code. Just structure it differently.

As it is, the code is often neglected. And rarely shared. Both bad facets.

And making a full text book that is executable is what I was comparing. That entire book is a program. I'd imagine that was much harder than the typical paper. Many of which are surveys and gloss over deficiency of the supporting code. (Again, I still concede the black swan nature of that book.)