Coding with agents is really good but I feel so empty

You ever beat a game and then unlock gun/sword/magic power that makes everything super easy to destroy and then you get bored of the game?

I’ve been adopting using AI agents at work to help ship features.

We have a pretty established codebase with very strict linting, formatting testing and e2e tests, so building apps with [the company I work for’s open source coding agent] has become a breeze and pretty safe to do.

It feels almost unreal how much I can deliver with these things. I just lay out an architecture plan, roll out segment by segment, code review every step of the way, I never try to let it write so much code it becomes a slog to review either. It’s way faster than if I started to type the code, and even when there’s architecture, system or just logical errors, it’s still faster to tell the cli to fix it.

Efficient? yes.

Shipping way more than before? Yes.

But it really doesn’t feel like coding. It's a bit boring...not that fun.

I mean, it isn’t coding. It’s spec definition writing and then code reviewing.

Two things, that are sometimes fun (spec defs, code reviews no) but they’re not coding.

I feel when AI autocomplete came out, a la github, coding became faster, exciting, and it still felt like my code; coding agents on the other hand feels like I’ve become a PM who’s really good at code review. It’s really a weird experience, I worry that new engineers will never gain the skills of actually knowing how these things lace together because they’re no longer doing, they’re just observing and referring (or just not reviewing at all). I’m sure I’ll also lose my touch if I do this too much.

I do feel like I’m still doing work though, there’s lots of code review, prompting and organizing still done, which for some people could be fun; but for me, I think I rather just start typing a for loop by hand sometimes again.