It is not a taste engine — it is an execution engine.

What this essay will cover.

Working notes from the UNSW guest lecture, being expanded into a full piece. The argument outline:

• The category error — when operators ask the model "is this good?" they're testing taste; the model is answering with execution.
• Why the model's execution is genuinely transformative — token‑per‑dollar economics, judgement‑free completion, infinite patience for the boring half of the work.
• Why its taste is unreliable — taste is calibration to a specific audience, not a universal property; the model averages where you need precision.
• The practical implication: stop asking the model "is this any good?" and start asking it to execute against a target you've already calibrated yourself.
• How this changes hiring — taste is now the senior contribution; execution is no longer a moat.

— Full essay coming. Reach out if you want the early version.