Generative models like GPT‑4o, Claude 3, Midjourney 6 and every task‑specific agent built on top of them have collapsed the cost of creation. Code that once took a sprint ships in hours. Blog‑posts, ads and playlists flow from a single prompt. We have crossed from scarcity of output to super‑abundance of output.
When the marginal cost of making something trends toward zero, the marginal cost of publishing trends the same. The inevitable result is déjà‑vu content—indistinguishable articles, cloned product UIs, cookie‑cutter code. Pat Grady called it the tremendous sucking sound of the market: an internet filled with AI slop.
⚠️ Signal‑to‑noise plummets. Discovery fatigue rises. Trust erodes.

If creation is easy, selection is hard. Taste—your practiced ability to choose, refine and elevate—becomes the scarce asset. It is:
| Practice | Why it Works |
|---|---|
| Study the Masters | Reverse‑engineer why great products feel great. Pattern‑match quality. |
| Cross‑pollinate | Bring design into dev, philosophy into product. Taste thrives on adjacent possibles. |
| Ruthless Curation | Keep a “hell‑yes” bar. Delete the rest. |
| Continuously User‑Test | Taste without empathy drifts into vanity. Ship, watch, listen, iterate. |
| Codify Principles | Write down the non‑negotiables. They become the guard‑rails for you and your models. |
Move from “Can we ship this?” → to “Is this worth someone’s precious attention?”
AI has inverted the economics of making things. The scarcest resource is no longer compute or even talent—it’s tasteful judgment. In the agent economy that’s forming, the winners will be those who wield models and maintain a razor‑sharp filter for what deserves to exist.
So build with abundance, publish with taste, and your work will stand out long after the flood of AI slop has receded.
Questions, counter‑points, or examples of taste‑driven products? I’d love to hear from you!