He was half crazy. When the transistor gave way to the integrated circuit computers of the day were built from the cheapest versions of these, the L series. You might need schottky ("S") parts for some critical high speed things but that was all.
There were faster parts, they were just expensive. That was not Seymour's problem. A cray 1 for example is built from 4 different chips, that's all he used. Really fast ones though.
Anything Seymour did used the fastest parts the planet could make. And then he didn't design some simple machine that executed instructions like a Z-80 or x86, his things were optimized for n-way fast floating point like you'd use for Fortran matrix math which is in fact mostly what they were used for. I'd hate to program a web page in Fortran but then I'd hate to have to program matrix multiplication in any languages other than Fortran!
So, Seymour considered it "done" if he could write a diagnostic in assembler than did floating point instructions faster than anything in the world. The rest was just software and he did hardware.
"But what about an operating system" - reporter
"Yes that would be a good idea. Why don't you buy one and write what you need. Next question? - Seymour.
To him it was an "Afterthought" and in his world if you wanted to do a weather simulation or DNA analysis you took your algorithm, translated it into machine language - which you had to do in a way to keep all the machines pipelines running at full speed - and how much of a big deal was it to have 12 lines of code a the beginning that read in the deck of cards containing the program? How much OS do you need in that context?
Alas it was the idea that something this powerful should be shared, and when it was idle that was a waste given the power and environmental support these things needed.
So the idea of sharing computers like this took hold. Atlas in England had enabled this with the first "multiprocessing" computer which mad the idea of computer-time "sharing" to be possible.
So a big committee was formed to define what software was needed to do this and was called Multics. This committee produced a spec and was so bad on paper a simple swapfile worst case might take half an hour.
Some guys at Bell Labs simplified this to the parts you actually need and coded it in a couple of months, the wrote a computer language to make that task easier and added self documenting features.
When a terminal in Bell labs went:
They were done.
So, in no small way Seymour's brilliance and arrogance (which I think was just fear. Good software IS scary) begat Dennis & Ken to create C/Unix. If Seymour had knocked something shitty out in a weekend like MSDOS and it became established then there's a chance Unix would have never happened.
Seymour Cray was to software was Dennis Ritchie was to computer hardware. Didn't know a thing about it, hated it. It was years before one of Seymours machines actually shipped with in OS. They're all sorta useless now. Whereas Dennis and Ken make your phone work. If you let Seymour do it cell phones would be an option that came with your car instead of a trunk. He was sort of the Rain Man of computer hardware and his computers made programming more difficult and had obstacles like 1) toggle in a bootloader via bit switches. It was a very special purpose tool. unix otoh was to be as general as possible and the hardware it was done on was not only easier but lent itself well to clever programming tricks (auto increment was born here, stacks were perfected here) that enabled the concept of Multics, an impossible task to actually be feasible as unix. That is unix could never have been born on a cray, it would have been a stretch to get ms-dos working on the thing and I wouldn't even try without something like node.js today. It's for doing matrix math and that's about all and your code just keeps stuffing numbers into pipelines and keeping them full. NASTRAN is a typical application.
Richard Sexton commented on Jonathan Zittrain's post. Funny thing is I've never used AV software my entire life. Never. In all that time my kids once put a virus on one of my machines and my boss did also in 1989. Other than that, da nada. I'm just careful. Don't click on programs to look at data, run trusted programs to open data files instead, you can't pass the program counter off to malware that way. And I don't so html email. That's just asking for it. And this is with windows, unix where I spend far more time doesn't really have this problem. So I just giggle when I see people get problems from av software.