Taste and Judgment in the Age of AI and LLMs

AI and LLMs have revolutionized content creation, making competent output cheap and accessible. This shift has elevated the importance of taste and judgment in technology. While AI can generate polished drafts quickly, the real value lies in discerning what is generic and what is truly valuable. Taste, in this context, is about making distinctions under uncertainty and involves noticing, rejecting, and precisely diagnosing what feels off. AI's pattern-compression capabilities often lead to outputs that are statistically plausible but lack specificity, resulting in a crowded middle ground of mediocrity.

The new bottleneck in this AI-driven world is judgment. The ability to critique and refuse generic outputs is now more valuable than mere generation. AI serves as a mirror for one's own taste, revealing the clarity of judgment. By generating multiple versions of a piece, AI helps refine taste through critique and iteration, but it cannot replace the human role of authorship and ownership.

Humans excel where AI cannot: holding stakes, working with new ideas, and making directional decisions. These tasks involve real-world consequences that AI cannot grasp. Builders must use AI to explore possibilities quickly but must also add context, constraints, and a personal touch to create meaningful work.

The risk is that AI could lead to globally polished but contextually shallow products. To avoid this, builders should focus on specificity, writing for real understanding, and designing for practical environments. AI should be used to explore and understand the design space, but human judgment must guide the final product.

Ultimately, taste is a byproduct of serious work and attention to reality. It grows through studying strong work, generating options, diagnosing failures, and shipping real-world solutions. AI makes the first draft easy, but it doesn't replace the need for human judgment, ownership, and the courage to create something unique. The real edge lies in using AI to eliminate average output and applying human judgment to direct, specify, and build beyond the statistical middle.