Vertical AI is compressing the B2C-to-infrastructure arc into a single move.

This is how we think of things for Shelf.

In prior cycles, consumer and infrastructure evolved sequentially. Google and Amazon built massive consumer audiences first, then over years exposed capabilities as infrastructure. AWS didn't follow S3 overnight. The cycles were long and linear.

What's happening now is different. Consumer and infrastructure aren't sequential anymore. They're simultaneous. Launch a consumer-facing tool, build a differentiated intelligence layer, then expose that layer as APIs and agent capabilities, all within the same product cycle.

Consumer builds data, UX iteration, brand, and real-world feedback loops. Infrastructure builds revenue durability, distribution hedge, and strategic leverage. They're not two businesses. They're one compounding engine.

Early examples: Mindtrip building an agent-first travel layer on top of its consumer planner. Daydream decoupling visual search intelligence from its consumer UI to power brand websites directly. Duckbill bringing human-in-the-loop trust into agent ecosystems via MCP.

The real question is whether there's a single intelligence core compounding across both surfaces, or just two adjacent products under one roof. The latter is operational complexity. The former is a control plane.

The strategic hedge: if distribution concentrates around a few dominant AI interfaces, you want to be the vertical brain they call. If distribution fragments, you want to own the daily habit directly. Building both surfaces covers either outcome.

h/t Kristen Green for this framing.