He puts words around what I've been feeling for the past 6 months or more:
> Ousterhout introduces the term “tactical tornado” for a programmer who churns out tons of working but overly complex code that doesn’t fit well into the existing system. Worse, a bad manager often sees a tactical tornado as the most productive programmer on a team, failing to recognize the complexity and code debt left behind for the rest of the team to clean up. If we aren’t careful managers, AI agents can easily become the ultimate tactical tornadoes.
Russ also has a great blog, with the most recent entry on fast floating-point printing and parsing. I believe these performance improvements have landed in the latest version of Go.
https://research.swtch.com/fp
I recently learned that Russ also wrote software for The On-Line Encyclopedia of Integer Sequences (https://oeis.org) and is the president of the OEIS Foundation.
"It used to be that if you could get a basic operating system coded and running, that was an indirect proof that you understood the basics of how operating systems worked."
How many -- and we are even supposedly talking about students -- have coded an operating system and got it running?
The bar used to be pretty low. My father spent a few years writing and selling a system/360 time sharing system in the 70s or so. He was not a self-identified computer person, just a physics student who found a way to make a few dollars.
In the 70s and 80s and 90s so many operating systems from lisp machines to ITS to the bsds... People used to write an OS like people write frontend frameworks now.
I think many good computer science programs offer a similar course where students make modifications to an open source Unix-like OS like Xv6, Minix, Linux, or one of the BSDs. I took such a course at Cal Poly San Luis Obispo in 2008.
However, I don’t know of any undergraduate programs where students are required to code an entire OS from the boot loader all the way up to basic drivers. Building an OS is part of the final chapter of From NAND to Tetris, so perhaps there might be some universities out there that do require their students to write an OS from scratch as part of the larger curriculum.
On my degree, we had to implement compilers, db system with B-Tree indexes mapping into i-nodes, extend an existing toy OS with cooperative multi-threadeding, and device drivers for PC floppies.
In digital circuits design, we got to create our own tiny CPU, and besides the digital design, some folks did the optional part of actually implementing it on a breadboard.
Among several other projects.
When a Software Engineering degree is 5 years, there is plenty of time for such stuff.
While not explicitly related to the topic any more: yes, things surely have changed, and I am not sure for the better. I suppose that was also what Cox was saying between the lines at least to some extent.
You can still write a simple OS on a modern computer, a lot of times its done using a virtual machine though to make iteration faster. It's not like rewriting linux or something, its just standing up the basics of machine abstraction.
Solid interview but honestly a bit boilerplate in parts. The bit about "program death" and Naur's theory is gold though, even worth re-reading every few years.
His take on AI feels a bit cautious though. I think he's underestimating how fast tactical tornadoes are gonna become the default just because managers love seeing shit get done fast, even if it's messy...
He puts words around what I've been feeling for the past 6 months or more:
> Ousterhout introduces the term “tactical tornado” for a programmer who churns out tons of working but overly complex code that doesn’t fit well into the existing system. Worse, a bad manager often sees a tactical tornado as the most productive programmer on a team, failing to recognize the complexity and code debt left behind for the rest of the team to clean up. If we aren’t careful managers, AI agents can easily become the ultimate tactical tornadoes.
Russ also has a great blog, with the most recent entry on fast floating-point printing and parsing. I believe these performance improvements have landed in the latest version of Go. https://research.swtch.com/fp
I recently learned that Russ also wrote software for The On-Line Encyclopedia of Integer Sequences (https://oeis.org) and is the president of the OEIS Foundation.
> Software engineering is what happens to programming when you add time and other people.
The book Software Engineering at Google goes into this (all factors of programming at scale / over time), worth a read.
It's available online for free: https://abseil.io/resources/swe-book
And, yep, it's worth a read.
Then a new term is warranted for what happens to programming when we add prompts & agents.
don't we already have the words slop and vibecode
An amazing distillation of the craft.
https://www.youtube.com/watch?v=hmq6veCFo0Y&list=PLrwpzH1_9u...
A good read. Except this weird remark:
"It used to be that if you could get a basic operating system coded and running, that was an indirect proof that you understood the basics of how operating systems worked."
How many -- and we are even supposedly talking about students -- have coded an operating system and got it running?
The bar used to be pretty low. My father spent a few years writing and selling a system/360 time sharing system in the 70s or so. He was not a self-identified computer person, just a physics student who found a way to make a few dollars.
In the 70s and 80s and 90s so many operating systems from lisp machines to ITS to the bsds... People used to write an OS like people write frontend frameworks now.
Russ Cox is one of the persons behind Xv6 (https://pdos.csail.mit.edu/6.828/2011/xv6.html), an x86 port of Version 6 Unix, which is the version of Unix described in the famous Lion’s book (https://en.wikipedia.org/wiki/A_Commentary_on_the_UNIX_Opera...). Xv6 was prepared for MIT’s 6.828, an undergraduate operating systems course. In this syllabus from 2018 (https://pdos.csail.mit.edu/6.828/2018/overview.html), students are expected to write significant subsystems in Xv6.
I think many good computer science programs offer a similar course where students make modifications to an open source Unix-like OS like Xv6, Minix, Linux, or one of the BSDs. I took such a course at Cal Poly San Luis Obispo in 2008.
However, I don’t know of any undergraduate programs where students are required to code an entire OS from the boot loader all the way up to basic drivers. Building an OS is part of the final chapter of From NAND to Tetris, so perhaps there might be some universities out there that do require their students to write an OS from scratch as part of the larger curriculum.
That's not weird? We did that in my OS class. Also even if you didn't the point still holds that the output was defacto validation of understanding.
I know of schools where that was part of the CS curriculum. I don't know if you couldn't graduate without it, but most people didn't.
> part of the CS curriculum
Well, maybe in the old days, when computers were simple (and even then, it would've been a very tough take in a curriculum).
Depends if you consider 1990's computers simple.
On my degree, we had to implement compilers, db system with B-Tree indexes mapping into i-nodes, extend an existing toy OS with cooperative multi-threadeding, and device drivers for PC floppies.
In digital circuits design, we got to create our own tiny CPU, and besides the digital design, some folks did the optional part of actually implementing it on a breadboard.
Among several other projects.
When a Software Engineering degree is 5 years, there is plenty of time for such stuff.
While not explicitly related to the topic any more: yes, things surely have changed, and I am not sure for the better. I suppose that was also what Cox was saying between the lines at least to some extent.
You can still write a simple OS on a modern computer, a lot of times its done using a virtual machine though to make iteration faster. It's not like rewriting linux or something, its just standing up the basics of machine abstraction.
I love the "Tornado" term and "software debt"... worked with a few over the years, fixed the bug but broke the paradigm. Well articulated wisdom.
Cloudflare says I'm not allowed to view the page. At least they are protected from "attacks."
Solid interview but honestly a bit boilerplate in parts. The bit about "program death" and Naur's theory is gold though, even worth re-reading every few years. His take on AI feels a bit cautious though. I think he's underestimating how fast tactical tornadoes are gonna become the default just because managers love seeing shit get done fast, even if it's messy...
Yet he is actively making the world worse at Google. Go (lol) figure!