3) The difference between something that looks plausibly like it might work when written down and something that actually works can be hundreds of millions of dollars of bugfixing.
There are a few patents that I've found informative and useful to read: Texas Instruments' calculator patents from the 1970s, the ENIAC patent, Astrolite, and Autotune. These are the rare patents where they actually explain things in detail, it reads like it was mostly written by an engineer rather than lawyers, and I know more after reading the patent than before. (Just want to give some recognition to the rare non-awful patent.)
Edit: I looked at a few of the Holmes patents. I'd put them in the awful category. Hundreds of pages of worthless discussion, followed by claims that are semi-incomprehensible, seem obvious, and have little to do with the earlier hundreds of pages. They have lots of nice graphs and diagrams though.
Eric Drexler's solar-sail patent was the only actually-educational patent I've ever seen, myself. https://patents.google.com/patent/US4614319 (Hm, I remember it as more detailed -- maybe I was remembering a tech report instead. But still this looks like reasonable English.)
I second that. In a previous life, I had to read and understand a few dozens of patents. Imagine page after page of buzzwords, unjustified claims, approximate logic, vague and meaningless sentences, clearly designed to be write-only.
In my experience, reading cutting-edge research papers in domains with which I'm barely familiar is so much easier.
There is a structural reason for this. What's the easiest patent to read? Show an object, name it clearly, describe what it does and how it's new. Easy!
But that is a poorly-written patent. As a competitor you can just read it, do the same thing but substitute something equivalent for one part, and you've just stolen the research investment that led to it.
A good patent has to describe the broad category (the innovation). And suddenly you go from clearly describing an example to using weird high-level vocabulary.
2) Patents are really hard to read
3) The difference between something that looks plausibly like it might work when written down and something that actually works can be hundreds of millions of dollars of bugfixing.