Taste used to be a byproduct. Now it's a curriculum.

One sentence this week did more work than most essays.

The reps were the curriculum, and nobody knew it

Addy Osmani, in Earning taste and judgment: "Taste used to be a byproduct of the reps. Agents took the reps."

Sit with the word byproduct. Nobody ever designed the apprenticeship. No one decided that a junior engineer would develop judgment by writing a few hundred mediocre functions and having most of them corrected. It happened as a side effect of the work being slow.

Which is exactly why removing it registered as pure gain. You don't defend a line item you never wrote down.

So this is not a story about juniors getting worse. It's a story about a curriculum that was never on the syllabus getting cancelled, and the people responsible for the syllabus not noticing for a year. The output looks fine. The output has always looked fine — that's what makes it a debt rather than an incident.

Here's the uncomfortable part for anyone running a team right now. If taste used to arrive free, and it doesn't anymore, then teaching it is new work that landed on your desk without a headcount, a budget line, or a name. It will not schedule itself.

This week's reads

Junior engineers are skipping straight to architect-level thinking The field evidence for the same claim, from the other end: juniors now arrive fluent in system-level framing without the years underneath it. LeadDev's conclusion is that apprenticeship needs a redesign, and I think that's right but understated. You can't redesign a thing you never designed. Someone has to write the first version.

Sensemaking as the Heart of Expertise Commoncog makes the case that sensemaking — reading an ambiguous situation and knowing what it is — can be trained directly rather than waited out. This is the piece that turns "teach judgment on purpose" from a slogan into an agenda. If you only read one thing here, read this one, because it names the thing you'd actually be teaching.

The Promotion Paradox Rands on your best engineer becoming the interim manager of their own team, and the question underneath it: do you actually want this job? Put it next to everything above and the trap shows up. The person most qualified to teach judgment is the one you just promoted out of the practice that made them credible. I don't have a resolution for that and I'd distrust anyone who offered one quickly.

Quick mentions

Engineering managers who build are pulling ahead — the counter-move to the paradox above: keep your hands in the work on purpose
Your junior engineers don't need an office. They need you — mentorship as designed structure, not proximity and luck
AI 'aha' team meetings — Lara Hogan on vicarious learning; the cheapest real practice in this whole issue
Scaling technical consistency — Will Larson on the same loss at org scale, where the shared memory that made decisions consistent doesn't survive growth
When you should delegate to AI (and when you shouldn't) — delegate less, own more

One ask: if you've deliberately put reps back in for someone on your team — real ones, not a training module — I want to hear what you did. Reply. I read everything.

And if this was useful, forward it to someone who'd argue with it.

— Wayne