As AI makes generic output nearly free, the advantage shifts to what AI cannot cheaply supply. Most people use AI lazily, accepting its generic results without adding quality or taste. Shann Holmberg argues that those who stand out over the next year will build three moats: - **Good taste**: the judgment to reject generic output and shape something better. - **Smart distribution**: reaching an audience, not merely producing. - **Strong agency**: the drive to improve and push the work forward rather than accept the first result. The claim is time-bounded. As AI gets better and more common, these moats compound for whoever starts building them now, widening the gap against those who accept default output. ## Related - [[LLMs Return the Average So Value Shifts to the Differentiated]] — the underlying mechanism: when the average answer is free, value shifts to whatever is non-average, which is what all three moats supply - [[Rejecting Good-Enough Work Is Taste Not Cruelty]] — the taste moat concretized: taste is precisely the judgment to reject the generic output most people accept - [[Selling Is the Scarce Skill When Building Is Commoditized]] — the distribution moat's twin: when producing is cheap, reaching and selling to an audience is the scarce complement - [[AI Age Preparation Framework]] — Dan Koe's independently-arrived-at parallel thesis with a longer 36-month window instead of this note's 12 months, same taste-plus-agency core --- *Source: "Everyone using AI has about 12 months to develop these 3 moats" — Shann Holmberg (@shannholmberg), X, Mar 2026 — https://x.com/shannholmberg/status/2031049690175652235*