Translator, not typist

I don't need to speak a language if I have a translator. That sentence used to feel like a confession — the generalist's excuse for not being the best SQL writer or the fastest front-end dev in the room. It's become the most productive belief I hold.

Here's what it doesn't mean: it doesn't mean the theory is optional. The opposite. When AI writes the first draft of everything, knowing what right looks like becomes the entire job. I can catch a wrong join in a dimensional model because I know the Kimball method — not because I hand-typed ten thousand joins. I can reject a lazy layout because I spent a decade building design systems. The translator handles the syntax; someone still has to know what the sentence should say.

AI amplifies judgment. It cannot replace it — and it ruthlessly exposes its absence.

The working loop is editorial, not mechanical. Intent first: what should exist and why. Then a tight brief — the same skill as writing a good ticket, except the recipient reads instantly and never gets tired of revisions. Then the review, which is where the actual craft lives: interrogate the output, test the edges, throw away the plausible-but-wrong. Ship only what you can defend.

What falls out of this is a strange new shape of person: one who can hold the customer problem, the system design, and the implementation in the same head, and ship all three without a handoff. Teams used to need that shape to be five people, and the five people leaked intent at every seam. The seams are gone. The judgment isn't.

So I've stopped apologizing for the translator. Syntax was always the depreciating asset; it just depreciated all at once. Theory, taste, and knowing where the bodies are buried in a data model — those compound. I'm investing where the compounding is.