Nate Meyvis

What software knowledge will stay relevant?

I recently wrote that a lot of our current first-order knowledge won't matter very much any more, but some of it will matter more than ever." I've been asked whether I know what that knowledge is.

No! I do not. Some of my initial conjectures didn't make much sense:

  1. I briefly thought that knowing stuff about distributed version control (e.g., git) would be durably useful, but now I'm not sure. I don't know what tools we'll use for version control (Jujutsu? Something else? People are speculating!). Moreover, I find myself delegating more and more git mechanics to AI.
  2. That situation is part of a more general pattern: knowledge of fundamentals, especially of fundamentals that tended to confuse programmers, attracted writing. LLMs thus know a lot about git, package.json conventions, and so on. If I suspect something is too essential not to know, they might actually be so essential that LLMs have had wonderful educations about them.
  3. As I've often written, I'm skeptical of my (or anyone's) ability to know much about what the future of software development looks like. I therefore can't know what first-order basic knowledge will remain essential.

That said, I'm often using my intuition for what choices cause observed behavior of a system. I'm often reasoning from what I see to "aha, X is probably depending on Y" or "why is it hitting the DB now?!" or whatever else. This isn't what I'm calling "first-order knowledge," but you can train yourself in it, and I expect it will be useful for a long time to come.

#future of work #generative AI #software