The best engineer in any larger org is rarely the most original one. They are the one who recognized that the problem in front of them was already solved by another team three quarters ago, copied the pattern, and shipped on Tuesday. Nobody in the all-hands celebrates this. It's boring. They celebrate the engineer who wrote something new.
The incentive layer is upside down. Reuse compresses cycle time, reduces variance, and produces fewer maintenance surprises. Greenfield work feels heroic and is statistically worse. Yet we celebrate "what did you create" and not "what did you avoid creating by finding it already done".
Agentic coding makes this worse, fast. Greenfield used to be expensive enough that someone in the chain would push back: "do we really need to build this from scratch?" Now the answer is "yes, easily, by Friday". The friction that used to enforce reuse is gone. Every team can spin up its own pattern, its own abstraction, its own slightly-different version of something that already exists in the codebase three times. The output looks productive. The org is silently fragmenting. Your codebase too. We are seeing a noticeable spike in microservices that boot huge frameworks, consume serious memory, carry full operational overhead, and exist to do one or two trivial things. Each one looked like the obvious build to the team that wrote it.
The bill comes due in two years. Six versions of the same thing, none of them documented, all of them maintained. Complexity scales linearly with new work and barely at all with reuse. The faster greenfield gets, the steeper the curve. Luckily we have AI to help us deal with it. The same tool that built the mess will be sold back to us to manage it. Another way to prevent fragmentation would be to include a mandatory check for prior art in the existing codebase. This could be easily solved by a custom skill.
If you want your engineering org to go faster, celebrating originality might actually hurt your progress. I think it makes sense to at least not forget to reward the engineer who closed their own ticket by pointing at someone else's PR from last year. That is relevant work. It should be seen as a success.
Thoughts? Find me on Bluesky.