AI makes us more of who we are - Ricky Smith

And it makes the wrong people productive

AI didn't make bad engineers good. It made them fast. It doesn't change who we are, it makes us more of who we are. It's a force multiplier and I don't mean that in a good way.

Used to be, a lazy or sloppy coder was self-limiting. They moved slowly. By nature of being lazy, they weren't motivated to make huge changes. Their code and structures may have been poor, but so, too, was their productivity. The blast radius of their bad decisions stayed small. Now with AI agents (and management breathing down our neck to use them) the same person ships 10x the code with the same judgment, same blind spots — just compounding faster.

It's the great equalizer, but in the worst way: it raises output regardless of value.

I see this at work with some of our large, shared repositories – like in a central repository that manages IAM roles for kubernetes workloads and another that manages service monitoring code for observability. These repositories and pipelines are centraly owned, but the code ownership is decentralized; fragmented by a directory structure. When a team needs to add their service, the common approach is to find an existing service's code, copy it, and make the minimum changes to get running. Nobody stops to ask if that module was well-designed in the first place — they just see a pattern and copy it. This has gone on for years. And the longer it goes on, the more confidence everyone has that it must be the right approach — after all, that's how everyone does it. Nobody notices they're working from forks-of-forks, and that the original source might never have been well thought out at all.

AI slots right into that workflow. Point it at a messy shared repo and ask it to add a service, and it'll do exactly what the lazy engineer did: pattern-match to the nearest existing example and copy it. It doesn't ask "is this module actually good?" — it asks "what's already here?" Like a human taking the path of least resistance, it defaults to the path of least resistance over the path of least entropy. The difference is a human might feel a flicker of guilt about it. AI just confidently ships the copy. And so, AI doesn't just speed up lazy engineers, it behaves like one — it avoids breaking bad patterns exactly like the engineer would. It perpetuates the tech debt.

So you get a double whammy: bad engineers move faster without getting better, and the tool meant to fix things has a default tendency to calcify whatever mess already exists — at scale, with confidence, and without the self-doubt that might've made a human pause.

Ricky typed this. An LLM was used for editing.