For most of history, the hard part of making something was the making. Writing the report, drafting the design, building the system: production was the bottleneck, and people who could produce well and quickly were valuable because of it. That is changing faster than most of us have absorbed. An AI agent can now generate a hundred competent drafts, a dozen plausible architectures, or a stack of reasonable designs in an afternoon. Production is becoming cheap. What is not becoming cheap is knowing which of those hundred is actually good.

This is taste, and I think it is quietly becoming the most important professional skill there is. When generation is abundant, value moves to selection: the ability to look at ten options and recognise that the seventh is the one, and to say why. It is the ability to notice that a technically correct answer is somehow wrong, or that a piece of writing is fluent but empty. And the uncomfortable thing about the current moment is that many people are responding to AI by producing more, when the scarce thing is the judgement to produce less, and better.

The good news is that taste is not innate, but it is built in two ways: deliberate exposure to excellent work: reading the best papers rather than the most recent, studying the systems that aged well, spending time with people whose standards are higher than yours. The second is articulating why something is good. Not "I like it", but what specifically makes it work. That articulation is what turns vague preference into transferable judgement. So, as a reflection for today: what are you doing to sharpen your taste, rather than just increasing your output?