I spent a week running Gemma locally, comparing it against Llama, Qwen and GPT-4o, trying to answer one question: Is Google's open model strategy finally good enough for real production workloads?
Over time, you learn that clean code is an internal concern, but stable interfaces are external. Users do not care how your system is structured. They care that it behaves the same way today as it did yesterday.
Founders are right to move fast. Early on, speed is survival. But some shortcuts do more than save time they quietly borrow against your future ability to build, change, and scale
I do not think cleverness is inherently bad. Some of it is necessary. Some of it is brilliant. Some of it is the reason a system survives real scale.
But none of it is free.