To pose it another way: for general software Dev (not some niche use case that only one guy understands), let’s say building websites in Vue or React, the idea of a 10x developer is that they can work in January and another, average, programmer can then spend February, March, April, May, June, July, August, and Sept… and the average guy won’t get more done?
Or they come in only every other Monday and contribute as much as a median Dev working full time?
It just doesn’t seem reasonable at all, at least if we read it literally.
I have worked with and met great, outstanding, developers who are much better than me. Some of them famous. None of them are anything close to 10x as productive, and I’m not particularly great. They’re much smarter than me, for sure, but they’re not 10x as productive.
How long would it have taken you or me to write Doom, or Quake? How long would it have taken Carmack, if it'd just been him?
I think you have a false dichotomy here. Carmack isn't 10x better a C or assembly than me. But in his domain he's easily 10x smarter and more productive than me.
And on top of that, he has a quarter century more experience in graphics programming than me. That shit accumulates.
Yes, actually. In graphics programming, from 2D and software rendered, to OpenGL and VR, you could give him and me (and presumably you) any task achievable in a few human decades, and he'd do it more than 10x faster.
"Graphics programming" is not a small niche, and yes in that space he's a 10x programmer.
I think you're thinking of this the wrong way. I agree that for something as simple as "build some websites", yeah, 10x is going to be very hard. You're then just turning specs into code, pretty much.
10x programmers are those that don't waste 9 months. The hard part is knowing which 9 months out of 10, which 9 ideas and components out of 10, is waste.
For "build a web forum" or something, yeah there are no 10x programmers. Though there are many many 0.1x programmers still.
> How long would it have taken you or me to write Doom, or Quake? How long would it have taken Carmack, if it'd just been him?
I'm not a games programmer. I doubt a median games programmer would take 10x the length of time that it would take any other games programmer. It just seems implausible, and I don't know why people want to defend it.
I spent 8 years at EA and was a senior software engineer. I have absolutely witnessed 10x game programmers.
I think in any field where you have a deep domain with complex, difficult constraints, it is entirely possible to have programmers with 10x the effectiveness of the median simply because of their greater domain expertise. They know the unique algorithms, data structures, design patterns, hacks, tricks, hardware quirks, tools, debuggers, code smells, etc. All of that has a huge compounding factor when it comes to output.
Here's a story I like to tell:
I joined EA shortly after the PS2 came out and worked at the studio that made Madden. Madden was a launch title on the PS2 (meaning it shipped right as the PS2 hardware did) and that version essentially sealed Madden's success. It was hugely successful.
But making a launch title is really hard. You're developing softare for a hardware platform that is itself in flux. When EA first started working on Madden PS2, they didn't have any dev kits to work on. I don't know if dev kits even existed. They needed to write code that they couldn't compile and run. But if they didn't start immediately, they would never be able to get the game done in time for the launch.
They had a guy on the team with a reputation for being a wizard. He took the MIPS R5900 reference manual, and disappeared into an office with blacked out windows for some number of weeks. When he re-emerged, he had written a PS2 emulator despite never having access to an actual PS2. The rest of the team were able to compile and run the game on that emulator so that they could make progress until eventually real dev kits arrived. From what I heard, when the real hardware showed up, the game actually ran on it.
Now, some of this legend may have grown in the retelling (it was already legendary by the time it was told to me), but I worked with that engineer a few times and I can vouch for his incredible ability to get stuff done. I remember once when FIFA was having performance problems and couldn't get their framerate high enough to ship. He was called into to help and a few days later, got it up to a solid 30 FPS. This was even though the team's own engineers had been trying to make progress and he had never touched the code before.
I think people dislike the mythos of 10x engineers because it's interpreted as some sort of cognitive essentialism. But I think most of it is just really deep domain expertise. Instead of talking about "10x engineers", we should talk about 10x at what.
The median games programmer wouldn't be able to do it in 50 years. (assuming they were locked in a room for 50 years with that same hardware, not that they got to see state of the art progress and upgrade their machine)
The amount of innovation, from applied BSP, Zbuffer tradeoffs, light maps with interpolation, interleaved FPU corrections for texture mapping, multiplayer protocols (qtest used TCP), brush-levels, etc...
Before ID software did these things, they weren't checkboxes for what to do. After, sure. But the hard part of programming is not typing.
The median programmer would not have been able to do it before it had been completely irrelevant. The median games programmer could do it in 2005. In 1993-1996? Not a chance.
The main competitor to Quake was Duke Nukem, which technologically was just "doom and a half". The enemies were sprites, ffs.
ID software was years ahead of every other games developer.
The reason other games licensed the Quake engine wasn't that it was cheaper than doing it yourself, it was that they couldn't do it.
Edit: granted, I have a low opinion of the median programmer. But sometimes in life you meet people who, with things you think you're actually pretty good at, just completely mop the floor with you without even trying (or showing off). It's humbling.
> I doubt a median games programmer would take 10x the length of time that it would take any other games programmer. It just seems implausible, and I don't know why people want to defend it.
Then why did all those thousands of other game programmers fail at creating a decent 3d game, before Doom was released? It's not like Carmack was the only one trying.
I think it's more likely that we wouldn't have thought something like Doom to be possible until Carmack came along and squeezed oodles of math tricks and hacks into the game to let it run on hardware of the era. So it may as well have be an infinitely long development time for us.
If you wanted to write Doom, starting in 1993, and take 10 years, your most efficient strategy would probably be to do something else for the first nine of them and wait for the technology to change around you.
Almost certainly you would just plain have not been able to do it at all. Not in 1993-1996. Not without it already having been done once.
And if you had, it would have taken you more than 10 years to realize what the right thing is, and to solve all very complex problems that showed up on the way.
Quake doesn't even have any load times! Not even with the hardware available at the time.
There is no "10X" touch-typist because a) the performance limit is absolute & relatively low and b) the learning curve is such that it's easy to get to the limit and plateau. Now let's increase complexity - still typing, but you must translate the text from another language. Increase again - now you're translating poetry. Increase again - the poetry is original Shakespeare. You need to translate into contemporary English approachable to Gen Z, as a script for a TikTok short, referencing memes from current events. There is no doubt that some people will have a combination of talents - in language, typing, copywriting, built-up cultural knowledge, lived experiences - that will make them 10X in such jobs.
The more software development work approaches touch-typing, the less difference between 1x and 10X — strict Jira tickets in, garbage out, maybe slightly faster. Spec-work, and most grunt-work, is undifferentiated and does not benefit from a 10X person. The more your work approaches poetry - such as exercising user empathy, improving the architecture, considering UX, respecting performance constraints, understanding user's jobs to be done, and improving your implementation around that - the more creative space there is for 10X to reveal itself.
Ironically enough your example is flawed: there are documented 10x touch-typists.
The realities of the court system place high demands on the typists; the minimum required typing speed is already quite high: trained court reporter or closed captioner must write speeds of approximately 180, 200, and 225 words per minute (wpm) at very high accuracy in the categories of literary, jury charge, and testimony, respectively[1] - and some exceed the minimum and go for 300 wpm. Even better, the official record for American English [is] 375 wpm [2].
Compare that to the average typing speeds around 30 - 40 wpm [3] Granted, the court reporters use specialized input devices (stenotypes) - but hey, the same can be said about highly productive programmers, who use specialized development environments - and sometimes also specialized input devices.
You make a good point that hints at another often-undefined assumption in these discussions - 10X performance vs who, exactly? IMO, it should be vs "a typical peer in that profession". For court reporters, if "a typical peer" is a trained court reporter who averages ~200 wpm, and a superstar is at 300wpm, then it's a 1.5X gain. If a "typical peer" is a regular Joe, then 300wpm would be 10X - but this doesn't feel appropriate because we compare productivity of a "10x developer" vs a "typical developer", rather than vs "an average Facebook visitor".
Even with drudge ticket-punching work, there is a vast spectrum of productivity between somebody who is capable and conscientious enough to do it correctly the first time and the person who has to rework everything multiple times because they can't do it right the first time and continue making mistakes.
That definition is too narrow. How about this one: can someone work for 1 year and (consistently) produce 10x more value for their organization than the median engineer? This seems transparently true if you've worked in any large organization.
Or they come in only every other Monday and contribute as much as a median Dev working full time?
It just doesn’t seem reasonable at all, at least if we read it literally.
I have worked with and met great, outstanding, developers who are much better than me. Some of them famous. None of them are anything close to 10x as productive, and I’m not particularly great. They’re much smarter than me, for sure, but they’re not 10x as productive.