> But what if we've been wrong about what matters?
Still wrong, though. Does the final product solve the problem it set out to solve? Because that is what matters. Specs and code are both tools to achieve that. So while this article is a good start... if you think that AI has changed what matters, not only were you wrong about it before, but you still are.
The idea of just re-building a whole app via LLMs when changes are needed is cute. But then you are also going to have a whole new set of bugs and UX changes. For anything non-trivial, documentation will be wrong, training will be wrong, people's experience will shift and complaints will ensue. Sure-fire way to increase churn, which is a bad thing.
> But what if we've been wrong about what matters?
Still wrong, though. Does the final product solve the problem it set out to solve? Because that is what matters. Specs and code are both tools to achieve that. So while this article is a good start... if you think that AI has changed what matters, not only were you wrong about it before, but you still are.
The idea of just re-building a whole app via LLMs when changes are needed is cute. But then you are also going to have a whole new set of bugs and UX changes. For anything non-trivial, documentation will be wrong, training will be wrong, people's experience will shift and complaints will ensue. Sure-fire way to increase churn, which is a bad thing.