1x Programming

Has anyone published some kind of quantified, "hard" comparison/survey about the relative productivity of programmers?

It would be very interesting with a clean "we asked n programmers to solve this"-type problem, and then being able to compare the solutions.

Uh, did I just sound like I wanted to work in recruiting? #fail

I imagine the point would be that pg writes many controversial things - with some of which the author explicitly disagrees - but this is one thing on which they do not disagree.

Let's suppose, for the sake of argument, that Newton once wrote, "The sky would be most accurately described as applesauce." (The absurdity is there to suggest that you would not agree. Take something more mild but still controversial and I don't see how it's different.)

> I don't agree with everything that Newton said but F=ma.

I think this is one of those things where its a question of what you're optimizing for. I don't like boxing solutions into `10x` or `1x` or whatever. I think this misses the entire point. It also assumes that single software engineers can make or break a project or company, and this is rarely the real case, even if it feels like it is.

I think an often missed piece in the formula is systematic incentives. At a company, this really means culture and process. If you have a culture and process of incentivizing novel solutions and so called `10x` behaviors, you'll have lots of `10x` developers, systems tend to produce the outputs that you foster. If you have proper incentives for high throughput, you will get it. No one engineer is "born" better than the other, like stated in the article, but a culture that fosters making good choices as a whole will see results align. This is how you can 10x your teams, regardless of individuals.

I think another way to look at is this: you're likely not the Yankees, you're the Oakland A's. Don't optimize for knockouts, optimize for runs[0].

[0]: https://en.wikipedia.org/wiki/Moneyball

I’m beginning to feel as though there simply isn’t that much novel code to write anyways. I very very much agree that programming is not inherently valuable, and is merely just a tool; I say this even as someone who got into programming at a young age and enjoys it for the sake of it.

I think in some narrow domains I can write interesting, and perhaps even mildly novel code. But a huge majority of the time, working for smaller startups, or Google, or myself, the majority of code I’m writing is something that has already been done in another form but I can’t use it for some reason. Maybe it does similar things, but it explicitly doesn’t try to solve my particular version of a problem. Maybe the license doesn’t work. Maybe it’s almost perfect, but it’s buggy or slow. Or hell, maybe I’m being a meme programmer and rewriting perfectly good code in Go or Rust just because I want to have the same thing but integrated with my favorite languages and tools.

The world of open source also changes things. You most likely should not hire someone to write a TLS stack. It’s not worth it. Chances are, a good TLS stack is not a feature that will set you apart, if it differentiates you at all in any positive ways. If you want to do things that existing TLS stacks don’t, it can still make sense to hire an expert, if that’s your thing. But, most of the time, even then, it would probably make the most sense to soft fork or upstream changes to an existing robust TLS library. Because of open source, it makes exceedingly little sense to actually have one entity maintain their own novel bits unless they’re unique and related to how the product differentiates itself. And even then, in some cases, it makes more strategic sense to just open source your bits anyways, and make yourself the standard. Google releasing Kubernetes is a great example; you can probably find many projects in the open source world, like Cockroach, which are basically clones of Google things they only released in a limited form, so the rest of the world went and standardized on non-Google implementations, and the Google versions didn’t benefit from cross-collaboration and the generalization of internal hacks. This also rings true for Facebook and React.

I think most programmers, given enough time and initiative and effort, could build cool and novel software, but the fact is, we don’t need novel software most of the time. We need better boring software. More general, less buggy, faster. If that’s depressing, well, I guess so be it. But there’s also a lot of toilets to clean out there, and someone has to do that, too. I can’t complain if it doesn’t feel like my work is exactly writing my name into history.

It's a meme that says some people are better at visualizing and mentally manipulating 3D images in their heads, and it implies that those people can use the same brain hardware to accomplish amazing feats of engineering

I dunno if there's any truth to it at all.. I think most of programming is less of an inspired act of genius and more of a tradeskill like plumbing or roofing.

I'm pretty sure there are tasks out there that I could get done in a month even though they'd take a team of 10 people 10 months to do, but if I trained those same people they could probably come close to my performance.