Post

qdot
@buttplug.engineer

before agentic coding: *writes code* *doesn't release* *gets busy, reviews a month later* what the fuck who wrote this over-complicated shit after agentic coding: *has robot write code* *doesn't release* *gets busy, reviews a month later* what the fuck who wrote this over-complicated shit

97 reposts935 likes

LenS
@lcsamuelson57.bsky.social

I tried agentic coding for a noncritical project, producing a not too incorrect but verbose, baroque, inefficient result. I need more practice in guiding an entity that writes code like a hyperactive student... I have (misplaced?) optimism that the "big guys" are vetting their results properly.

qdot
@buttplug.engineer

It takes a ton of work to figure out how to steer things correctly, then make sure it all stays on the rails. Dunno if it's as much as writing the code yourself depending on the task, but it's a totally different type of energy used to do it.

Azure Wolf
@azuredawolf.bsky.social

Who would have thought we'd ever be feeling nostalgia for an era where we were reviewing non-agentic code? Or when the idea of an agent writing our code seemed like something WAY out into the future?

Funwayguy
@funwayguy.bsky.social

In QA, that shit happens every other day. We review oceans of code slop not even the devs can explain why the fuck it was written that way before rubber stamping their copilot commits.

qdot
@buttplug.engineer

ai haters: why are you using ai ai boosters: why are you insulting ai people who actually know ball: all code is horrible, none should ever be made in any way

qdot
@buttplug.engineer

Woke up to so many replies about how I’m either doing code wrong, ai wrong, posting wrong, or all of the above that I thought I’d accidentally opened masto

hare_ware
@hareware.xyz

Aren't you supposed to make a Reddit post claiming your brand new software revolutions everything (but it's a slight variation of an existing software but with less polish and crazy scope creep), and then get busy and forget about it?