It Was Never Taste. It Was an Unexplained Rule.

A LeadDev piece making the rounds this month argues that engineering managers need to get back to hands-on building because, in the author's words, "taste has never been more important." His reasoning: AI is designed to find the median of human taste, so telling the difference between what he calls "AI slop" and something genuinely good is a skill you can only develop by doing the work yourself. Nowhere in the piece is there a number, a team comparison, a single measurement. Just the assertion, and the word doing all the work underneath it.

I read pieces like this with real frustration, and it took me a minute to place why. It is not really about coding. It is that "taste" is being used, across an industry that should know better, as a polite way to say a skill cannot be taught, which conveniently means nobody has to try teaching it.

Conventional wisdom right now treats taste as the legitimate, almost untouchable differentiator for technical leaders in the AI era: proven only by doing, developed only through hands-on building, the mark of someone who has put in real hours rather than delegated everything away. A manager who has it is credible. A manager without it is just another person clicking through AI output.

Calling a judgment call taste makes an implicit claim: this cannot be broken down, it can only be developed by doing, so do not ask me to teach it. That claim does two things at once. It excuses the leader from the actual hard work of mentoring, which is breaking a felt judgment down into the pattern or test that produces it. It keeps the leader as the permanent bottleneck, because if taste genuinely cannot be taught, only the leader, or whoever already happens to have it, can be trusted to make the call, so every decision keeps routing through them indefinitely.

The leader saying it usually is not being cynical. He likely believes it. But belief does not change the effect: the team is left waiting for direction constantly, because nobody ever tried to name the actual test underneath the feeling. This is what I call The Halo: a leader's sincere helpfulness, genuinely meant, produces learned dependence instead of capability, and the leader is often the last person in the room to notice, because it looks like generosity from the inside. Most taste is a compressed heuristic that has never been unpacked, not a truly ineffable trait. The tell is whether anyone has actually tried to articulate it and failed, or never tried at all.

None of this means taste does not exist, or that everyone's judgment is equally good on day one. It means taste has to be a last resort, not a first answer. A leader owes the team a real attempt at naming the actual test underneath the feeling before concluding it cannot be taught, the same move that makes Second-Order Coaching possible at all: judgment distributed downward instead of concentrated in whoever already has it. If that attempt genuinely fails after real effort, say so explicitly. Reaching for the word before making the attempt is the default, not the exception.

Think of a judgment call you have made that you have never fully explained to anyone on your team, something you would chalk up to experience or feel if pressed. Try to name the actual test underneath it, right now, in one sentence. If you can, it was never taste. It was an unexplained rule. If you genuinely cannot, after actually trying, that is the first honest use of the word you have had in a while.

LeadershipOS™ has the full method for turning a felt judgment call into something a team can actually inherit, Second-Order Coaching instead of a word that ends the conversation: http://TheLeadershipOSBook.com