What I believe about the future of software development

I think there's a bifurcation playing out that LLMs are accelerating, and that these points will be true for a subset of us, not for everyone, and to some extent as self-fulfilling prophecies. If you think good code doesn't matter and performance rarely does, you're in a bubble - these aren't true for everyone.

Code review will die.

I definitely think a lot of the nitpicking that people did in code review is going to die, and that it will become somewhat unrecognizable to someone used to doing code review 5-10 years ago.

Unit tests might die too.

Kind of doubt it. Unit tests are the guardrails that keep the robots from making errors, just the same as they did for humans. You can now have lots more of them, and they're easier to maintain.

The craft of writing code will disappear.

Who will watch the watchmen?

And I'm still managing to take Claude's suggestions, implement them in a different way, and Claude agrees with me that I improved the design...

There’s no proof that “good code” will matter in the future.

A lot of "clean code" practices were garbage, and should have been thrown away long before the robots took over, but I think there's still going to be a problem where if you're feeding spaghetti into the robots, you're going to get bad output. And just asking the robots to "fix all the complexity in the code, don't make any mistakes" is unlikely to be satisfactory.

there will be people employed to write code in 10 years. But do you want to have that job?

Yes. Now go away, I'm trying to write some code.

I dunno somehow this doesn't land, if what you are making is an LLM coding tool -- which I believe Amp is

It's similar to Steve Yegge making proclamations about coding when he was also making LLM coding tools - SourceGraph, Gas Town, and then some new thing

It seems too circular. What about writing software with LLMs that itself has nothing to do with LLMs?

The craft of building software will be more important than ever. Knowing how to solve business problems with software, how other software did it and why and why not, when and how to ship it, how to get feedback on it – that’s the new game.

The new game? I have found in my career this is the only game that actually matters when it comes to software development within a business context.

There’s some insightful stuff here, but I do have some disagreements too.

Unit tests might die too.

I think that they are likely important for LLMs the same as for us: to prevent regressions.

Open source in its current form doesn’t make sense anymore. “Given enough eyeballs, all bugs are shallow” is still true but now we have artificial eyeballs.

I never thought free software was principally about free bug reporting and fixes: I always and still think that it’s about user freedom. LLMs enable users who are allowed to code but do not know how to do so: I think that they make free software even more relevant than before!

The terminal is dead.

Strange take, because command lines are a great way for agents to work, and terminals are great ways to work with agents. And of course Emacs is a grey way to work with agents, too, and it still works in both a GUI and in a terminal.

Moreover, there are plenty of non-coding terminal tasks, for example just moving files around. I don’t use a GUI file manager because the terminal is so much easier and more efficient.

There’s no proof that “good code” will matter in the future.

Seems like code which is more token-efficient will be better in that sense at least.

This was originally posted on X and blew up.

Well that's enough to immediately stop reading the rest of the post