AI ate the apprenticeship; taste is the last thing left on the plate

This essay argues that AI coding tools have eliminated the friction that once filtered software output and, crucially, taught developers taste. The cost of making things used to ration output and educate judgement; now that cost is gone, flooding the field with "good enough" slop. Taste—the wordless, hard-won ability to recognize quality before you can justify it—was always the non-mechanical core of the craft, and it still cannot be automated. But taste is slow, invisible on dashboards, and uncreditable, so the market rewards the tasteless fluent shipper equally. The author worries that removing friction removes the curriculum, producing developers who can make anything but cannot tell whether they should.