Redundant

I’m using AI to code. At least, I’m trying to.

I’ll take all the efficiency gains I can. I like coding, I’m doing it Sunday today. But I also like building software that actually works and is useful to somebody, and it currently takes a lot of work. I work out, I have a hobby, a girlfriend. Life’s short and I need that extra efficiency.

Yeah, maybe a year or two from now, the AI will make us knowledge workers redundant, and we’ll all be unemployed, or worse, Project Managers.

As of April 6th, 2025, it’s still kinda ass. It’s funny that it’s ass because a lot of its work is just redundant.

Very often I find myself asking Claude “Do you need this?”

I ask it for the simplest impl, and then I remove stuff from it. Comments that say nothing useful. Unfinished parts of features I didn’t ask for. Unused duplicate logic.

Like if I was working with a junior.

It reminds me of when I worked with Maciek, a great guy, but back then he was 19. He knew how to code. He was already a good engineer. He just didn’t know JavaScript.

I remember one code review when I said “Okay, this is good. It should work. Next time you can use spread syntax instead of a loop here.”

I was that kid. I asked “Where do I find the repo with tests?” on my first day (oh naivety). And “Can we gradually migrate this to Flow? We’d probably have fewer crashes on prod if we had static typing.”

I grew up, tho.

With experience, you acquire more than taste. You become more tired. You burn out. You get old. Experience doesn’t make you lazy. Or it does, but lazy like in lazy evaluation. “We’ll burn that bridge when we get to it” type.

We learn to do less things that are unimportant.

AI isn’t there yet. And removing unimportant cruft in future iterations is not the same as not adding it in the first place. Putting in more work to fix the prior work is lossy and prone to drift.

Iteration is great, but you gotta remember what you’re doing. Don’t just spin frivolously. This applies to humans and AIs.

Happiness

With humans, you can always try to explain the redundant work.

My father has a 3D printer and a dream he’ll make money on it. The silly little plastic dogs he prints are totally redundant, but the process isn’t. It makes him happy. Without it, he’d have to get to terms with the fact that his time has passed. You won’t convince a Polish 65yo to go to therapy, so this is a kind of therapy. A thing that fills his heart with joy, and gives him something to be excited about.

In the case of AI, the only explanation I could find is that the more tokens you spend on Chain of Thought, the slower it feels. So we have a trade-off. Can’t think too long and iterate in silence, without a human in the loop, because the human’s experience will degrade. At the same time, the AI should think just a bit instead of hallucinating useless slop.

The LLMs will surely get better with time, able to one-shot high quality stuff. Everybody will be able to build an app. Everybody will be able to generate pictures unrecog nizable from human-made.

I guess so, if it makes us happier.

But does it? YouTube and all the twitters are already dead internet, especially if the post is about politics. I hope the superintelligence comes before we get even more flooded with slop.

I just remembered a tweet from half a year ago.

Yeah, this fits.

I’d rather get unalived by liquid metal Robert Patrick than use Android again.