The Two-Rate Diagnostic

How long will your AI advantage last?

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.

Two steps · about a minute · nothing you type leaves your browser.

Built for the essay How Long Until Your AI Edge Stops Paying?

The clock starts the day your edge works
Step 1 of 2

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.

Step 2 of 2

Your two rates

One we judge from dated signals. One you have to measure on purpose.

Your absorption rate
Of the new AI capabilities you seriously tried this year (a copilot, a forecasting model, an automation you rolled out), how many moved a number your CFO or customer would recognise? A moved number is a renewal saved, a price you raised, hours or headcount out, revenue in. It doesn’t count if you only shipped a feature or the demo impressed people. Running a multi-unit org? Run it once per business unit; a blended number hides which unit is the bottleneck.
A rough count is fine. Round to the nearest few; the ratio is what matters.
Do you own the layer you named?
Answer for the layer that matters most to your business. You can run it again for another.
Add your two numbers above to see your result.
Your two-rate read

Your move
Start building your next layer within

Your one line

Your clock against your payback
Your absorption rate
Your layer’s half-life

One rate you measure: your own count of what landed. The other is our judgement of how long your layer stays scarce: each layer’s range was set by hand on 25 Jun 2026, and your answer to the clock question picks your end of it. No source states these months. The dated signals below were re-checked on 7 Oct 2026 so you can test the range against them; no range has moved since it was set.

Dated signals to test this clock against

Each links to its source. Re-checked 7 Oct 2026; some are newer than the range, which was set 25 Jun 2026.

Where you sit · absorption × the clock
Watch
Where value is still scarce

The full map

Seven layers a capability can pass through, each on its own clock, from the ones still worth owning to the ones already melting. Your layer is marked, at your end of its range.

Most moats are a stack
Own another layer too? Add it, and your floor becomes the slowest clock you own.

On the Shelf-Life Path? Day 2 is The Shape Test: is your growth actually compounding?

A free tool from The Durability Curve · companion to How Long Until Your AI Edge Stops Paying?
Ranges set 25 Jun 2026. Signals re-checked 7 Oct 2026; no range has moved since. The ranges are our judgement: no source states the months, and the dated, linked signals on each layer, some newer than the ranges, are there so you can test a range against the evidence. Falling price or margin at a layer is the cleanest sign its clock is running. The build window sits at a third to a half of the half-life, rounded out to whole quarters with a one-quarter minimum, on the premise that a build-and-adopt cycle runs one to two years. A layer with a half-life under 12 months is commoditising, and its window is this quarter. Calendar dates in your result count from today on your device. Paybacks are read in whole months, up to 240.