The Productivity Mirage
18 points by carlana
18 points by carlana
Yesss, I already use vanilla Sublime with half-broken syntax highlighting! I'm going to be sooo productive.
But how did he do it? It feels like this post should be just the introduction to a longer piece that examines how Bob leveraged product sense, but all it presents in the shock of seeing someone talented using Sublime Text.
I don't know, i still think some level of productivity setup helps. Maybe some folks need it more than others....but maybe engineer Bob might have simply been either very talented or perhaps the kind of person who can focus so greatly. (I suffer from getting distracted quite easily.) I do acknowledge that its indeed a slippery slope that a person might start to spend far too much time on their "productivity setup". But, i think like in other aspects of life, moderation is best.
I think the real moral of the story here is (or at least should be) not that customizing things is bad, but rather that you need to stay focused. For example, I run a pretty custom Neovim setup with a bunch of third-party plugins and a bunch of my own. However, I don't really change it much other than when some new Neovim version requires certain changes. The result is that, at least by my own judgement, I am very productive but without constantly going "oooh, let's try out new fuzzy finder #15".
i still think some level of productivity setup helps
Absolutely, but the point is
I'm very lucky to work with 2 very skilled people like the one described in the article. One uses stock Vim and the other one uses stock Emacs, both without any plugins. Bare minimum configuration on stock Ubuntu LTS. I ran Vim without syntax highlighting for some time and while in the long run it improved my productivity by some measurable (but small) margin, I obviously didn't become as skilled. I think that they can focus on the problem in a way that most people can't. They just sit on the chair and immediately get to work with 100% focus, then get up 8 hours later after some deep thinking every day. I still haven't figured out the secret formula.
The lack of existence proof for productivity outcomes is something that's kinda been bothering me for years.
Assuming there are better and worse ways to write software, better and worse tools, better and worse methodologies, better and worse languages; why doesn't this appear to be a competitive advantage for software companies? Why isn't rust, or haskell, or whatever language that claims to produce less defective code, why isn't it winning capitalism? Why is Java EE and SAFe ostensibly the winning combination instead?
why isn't it winning capitalism?
Because the experts at capitalism optimized for winning directly. Productivity, quality, customer satisfaction, or sometimes having a product at all (FSD, on Mars), are only tangentially related to making money, and can be skipped.
More seriously, success of languages depends on lots of factors, some of which are non-technical and unglamorous. We'll see how LLM rewrites change the picture, but before that languages had a lot of momentum and network effects (compatibility, ecosystem, userbase, being in curriculums, vendor backing) which kept them as the overall rational best choice long after they weren't best choice on technical merit or productivity measured in isolation.
There's a feedback loop to break out of: for software engineers, there isn't really a point in learning Rust if you want a job, because jobs are hard to find in it. And so the pool of rust devs does not grow very quickly, except for people who choose to learn Rust for the love of the game. And so companies are not incentivized to choose Rust, because the pool of hire-able candidates is small compared to people who know (or could learn) Java. I also don't think it's controversial to say that Rust is harder to learn than Java. I personally struggled quite a bit with learning Rust.
Some companies can get away with placing themselves in opposition to this, and use Rust to attract employees who are passionate enough to learn Rust on their own (I remember this being a thing with Haskell also). But this is not very scalable, and the sort of people who are willing to teach themselves Rust tend to be already employed anyway.
the sort of people who are willing to teach themselves Rust tend to be already employed anyway.
crying
But yes, most of that tracks.