When the cost of production is high, validation needs to happen at the planning stage; when the cost of production is low, it’s easier to just produce whatever thing needs to be tested (software, copy, creative, etc.) and see how it performs in the wild.
By lowering production costs, generative AI is allowing us to move validation downstream. In other words, we’re relying less on the map and more on the territory.
This is similar to how, in an ideal world, we would not have a candidate screening process. We would just hire them and see how they perform in the real world. This is not feasible, so we need to approximate that evaluation; but we should be mindful that it’s just that—an approximation.
One side effect is that this shifts us from generative toward reactive collaboration: when producing a complete artifact is cheap, it’s easier to generate one solo (with a model) and then solicit feedback than to co-create it from scratch.
This goes partially contrary to the idea that efficiency is not effectiveness or, as they typically say, “you don’t just need to move fast: you need to move fast in the right direction.” It turns out that moving fast is a way to accelerate the search for the right direction (though a balance is necessary).