Note
Are We Building the Answer, or Building to Find It?
Cheaper execution does not mean we need more certainty before we build. It can mean we need less certainty about the solution and more clarity about what we are trying to learn.
One of the things I like most about building with AI is how cheap it has become to change my mind.
A direction that once would have meant days of implementation can sometimes be tried, rejected and rebuilt before I have become particularly attached to it. That is liberating, because it means I do not always need to know the answer before I start.
That has changed how I think about clarity.
I used to frame the argument as: as execution gets cheaper, clarity gets more valuable. I still think that is true, but it is incomplete.
Cheaper execution also means we can afford to fail faster.
If something takes six months and £1m to build, you understandably spend a lot of time trying to make sure you are right before committing. If you can build enough of it in an afternoon to discover that an assumption was wrong, being wrong is suddenly much cheaper.
So the shift is not towards greater certainty before action. It is towards greater clarity about what we are trying to learn.
There is a big difference between building something because we believe it is the answer, and building something because it will help us find the answer.
In the second case, discovering that we were wrong can be a successful outcome. We thought users would behave one way and they did not. We thought a feature mattered and nobody used it. We thought a process needed automating and discovered the real problem was somewhere upstream.
If each discovery costs six months, they are expensive failures. If they cost an afternoon, they are useful information.
This does not make clarity less important. It changes where the clarity needs to sit.
We can be uncertain about the solution while being very clear about the problem, the hypothesis, the boundaries, the consequence of being wrong and what we want to learn.
That is one of the more important changes AI brings to development for me. It does not just let us produce known answers faster. It lets us use building itself as a way of finding better answers.
Failing faster is not necessarily a failure of AI-enabled development. Sometimes it is the point.
Are we building the answer, or are we building something to find the answer?