You Had No Taste Before AI (matthewsanabria.dev)

The piece argues that calls to “develop taste” for using AI are less a new requirement and more a reminder of basic creative and professional skills many people never cultivated. “Taste” is defined practically — contextual appropriateness, quality recognition, iterative refinement, and ethical boundaries — and the author says these skills predate AI. The problem isn’t AI creating poor content, it’s that AI amplifies whatever standards (or lack thereof) people already have. Those who preach taste often never demonstrated it before AI; those who succeed with AI already had strong judgment and simply applied it faster. For the AI/ML community the takeaway is operational: good AI outcomes depend on human curation, domain knowledge, and iteration, not mystical new competencies. Technical implications include the need for evaluative criteria to separate “useful” outputs from “slop,” tooling and workflows that support rapid iteration and provenance/ethics checks, and cross-domain fluency (breadth) to switch contexts effectively. Practical steps offered: compare a best and worst piece of your work to identify differences, study exemplary artifacts in your domain, and iterate on AI-generated outputs with targeted critiques. In short, invest in timeless standards of taste and process design — the medium changed, but the fundamentals remain the same.