How long does your layer stay scarce?
You can be right about AI and still get nothing for it. Value collects at the scarce layer every customer passes through. Name yours, and see where you sit: your absorption against your layer’s clock.
Name your absorbing layer
The thing a new AI capability has to pass through to reach your customer.Example: if a hospital won’t switch away from you, a better model doesn’t help a rival catch you, so that edge is real. If your edge is a clever prompt, a better model just erases it. Pick the layer your advantage actually runs through; if a few fit, the one hardest to copy. If you don’t sell an AI product but use AI to run your business, that’s usually the process you rebuilt around it: Proprietary workflow integration.
Your two rates
One you read off a press release. One you have to measure on purpose.
One rate you measure (your own count of what landed); one we estimate (our dated read of the layer’s clock, which you refine below). The plane pairs your arithmetic against our judgement, not two measured coordinates.
Every instrument is sourced and dated, like the one you just ran.
The full map
Every layer is on a clock. These are the seven a capability passes through, from the ones still worth owning to the ones already melting. Your layer is marked.
Snapshot current as of 25 Jun 2026; every signal dated and sourced. Falling price or margin at a layer is the cleanest sign its clock is running. The half-life ranges are our dated reads off the signals below; the specific months are our calibration off each layer’s ranking, not a figure any single source states, and we revise them as the signals move. The build-window sits at roughly a third-to-half of the half-life, on the premise that a build-and-adopt cycle runs one to two years.