The Risk and Reward of Agentic Coding
A lot of developers worry that agentic coding produces slop and messy codebases. For me it comes down to risk and reward. On a personal prototype the risk is close to nothing. At a startup I'd take more risk than a large company would, with eyes open.
A lot of developers are taking positions on agentic coding and so-called vibe coding, and a lot of them are worried. It produces slop, nobody's looking at it, the codebase is messy, the architecture doesn't scale.
There's some truth to that. I also suspect some of it is self-serving, maybe not on purpose, and that some of these opinions haven't kept up with how fast the models are improving.
For me it comes down to risk and reward. Maybe I have a messy codebase. So what? If I got it quickly and it does the right thing, that's a lot of so what. Does the mess hurt me more than the delay or expense it would otherwise take to have something functional in the first place?
Some of it depends on the organization. For my personal prototype experiments there's almost no risk. The reward is that maybe I make something cool. The risk is nothing, because nothing is what I was going to have anyway. That's my BATNA, my best alternative.
If I were at a large, high-risk company, I'd be more conservative for sure. At a startup, I figure we ought to take a little more risk, because if we don't, we're going to get eaten alive by a large company or by someone else who did.
I'm willing to take that risk. I do it with eyes open.
Last updated