Empirical Taste (ajaysquest.beehiiv.com)

Ajay argues that as AI automates creative work, “taste” becomes the last competitive moat and can be split into two kinds: epistemic taste (learned understanding from critique and context) and empirical taste (instinct honed by exposure, consequence, and repetition). Empirical taste is essentially pattern recognition with emotional confidence — the subconscious calibration that makes a choice “feel right” before you can explain it. Unlike mystical intuition, it’s measurable: it compounds through build→test→learn→adjust cycles and is roughly proportional to the number of feedback cycles survived. For the AI/ML community this reframes human judgment as a trainable signal rather than an irreducible oracle. Data and A/B tests don’t kill intuition — they become its training set, shifting weights in an internal model until practitioners “feel” the numbers. Practically, that means compressing feedback loops (startups beat slow corporate cadence), instrumenting experiments, and keeping human-in-the-loop workflows so designers and operators can translate gut feelings into actionable changes. Key implications: treat aesthetic calibration like model calibration, use rapid experiments as supervision for human decision models, preserve dashboards early while allowing trained instincts to guide rapid choices later, and value mileage (exposure) as a form of expertise.