Coding was never the bottleneck

Coding was the reward, at least for me.

See, running a business involves a lot of boring, repetitive crap.

You do marketing, accounting and payroll. You answer support tickets and talk to customers. And while doing that - boom - you spot a pattern. Different customers keep describing the same problem in different words. Eventually it clicks. You need to build a new feature that solves it for all of them. Brilliant, can we start coding now?

Nah, first you research possible solutions and check what competitors are doing. You think about pricing, tech debt, design and architecture. Also, will it make onboarding too complicated? How do we market it? Does it even fit our product positioning? Will it break backwards compatibility for the self-hosted customer still running version 0.0.432b in a basement somewhere?

Then finally, you can start coding.

An excuse to get back to my favorite thing. The 5% of the work that's actually fucking fun. I get to craft a perfect, scalable, elegant solution on the backend. Pixel-polish the UI on the front, make it fast and good-looking without bloating Tailwind's output.css to 150 KB.

AI took that away.

AI took that away

Which is funny, because writing code was almost never what slowed me down.

The slow part was deciding whether the feature should exist. What was the customer actually asking for? Would fixing one person's problem screw up another person's workflow? Was a minor improvement worth another setting we'd have to support for the next fifteen years?

If anything, fast code makes those decisions even more important. When implementation is expensive, bad ideas die naturally - nobody wants to spend three weeks building a feature they're not sure about. When implementation takes an afternoon, every half-assed thought suddenly looks affordable. Nice, you removed the coding bottleneck and replaced it with a maintenance firehose.

Constraints are good. They keep your plate small and force you to prioritize. This forces creativity.

"A change is as good as a rest" ™

Switching activities is also the best way to rest, it refreshes both the mind and the muscles. Coding used to be that break for me: after hours of exhausting business decisions, I could switch into a different mental mode and go build something. Same workday, different part of the brain. I felt fresh and got my daily dos of hard-earned dopamine.

Years ago I wrote that Coding is procrastination for founders. That was before AI. Now I'm stuck in an endless decision loop: research, prompt, review. Research, prompt, review. It's more efficient for sure. It's also one long, monotonous management task with no mental gear change.

I went from building the thing to supervising the building of the thing. Again, probably more efficient. Also suspiciously close to "management", a career path I spent years avoiding by starting my own company in the first place, ffs!

This doesn't mean AI coding is useless or overhyped. It is ridiculously useful, small teams can attempt projects they couldn't justify before. Old bugs get fixed, tedious migrations become possible. Sh*t, I'm not getting anywhere near "WhatsApp business API" (or whatever crappy big corp API) without AI ever again, never. The gap between "we should really fix this, someday.. maybe..." and actually fixing it has collapsed.

But code was never the product. Solving the right problem was the product.

The limited resource is still judgment and taste. Knowing which customer request is a clue and which one is just a person having a strange Tuesday.

Only now, when I finally reach the fun part, the computer says "I got this."

Well.

Thanks.

I guess.