# Emit Content Machines Can Consume Rather Than Embedding AI ## Claim A product earns more by emitting content that machines can consume cleanly than by embedding AI features inside itself. <!-- atomicity: 1 idea ✓ --> ## Scope & conditions - The emitted surface keeps working as models change, while an embedded feature is pinned to the model generation it shipped against. - Applies directly to courtneyr.dev and to post-kinds machine-readable output, where the leverage is in what the site publishes rather than what it runs. - Not either-or in principle, but they compete for the same build time, and the emission side is the cheaper and longer-lived half. - Weakens where the value genuinely requires inference at the point of use rather than clean structured output. ## Evidence From Dave Winer's post on Markdown: the argument is for building products AI can read rather than products containing AI. ## Links - [[A Format Whose Source Is Its Presentation Cannot Hide Bloat]] — the format-level property that makes clean emission possible - [[Notes Are Files Not Database Rows - Durability Beats Features]] — the same preference for a durable readable substrate over feature surface - [[3. Resources/MOCs/AI & the Future of Knowledge Work MOC]] — the publishing thread this product decision belongs to.