Public experiment · September 24, 2026
A quieter mobile shop bar, tested with ĀML.
On the Samples and Ridge Kit pages, the fixed mobile bar repeated a shortcut to the page a visitor was already viewing. A pinned ĀML browser compiler now evaluates those declared scenarios and hides only the duplicate current-page shortcut. The phone contact remains visible.
What changed
Before
The fixed mobile bar showed Samples, Build kit and Call to order on every route, including the destination pages themselves.
After
The current-page Samples or Build kit shortcut is suppressed. Other shortcuts and Call to order remain; the bar changes from three equal columns to two when one is hidden.
Compiler decisions
This table calls the same site adapter that controls the mobile bar. Values are project-declared design inputs, not measurements of visitor attention or benefit.
| Page | Samples | Build kit | Call to order |
|---|---|---|---|
| Samples page | SUPPRESSCost 3 · value 1 | ALLOWCost 2 · value 4 | ALLOWCost 1 · value 5 |
| Ridge Kit page | ALLOWCost 2 · value 4 | SUPPRESSCost 3 · value 1 | ALLOWCost 1 · value 5 |
| Putting Greens page | ALLOWCost 2 · value 4 | ALLOWCost 2 · value 4 | ALLOWCost 1 · value 5 |
Safety and scope
If the browser compiler fails or returns an incomplete decision set, all three navigation actions stay visible. Desktop navigation, page content, sample requests and phone ordering are unaffected. No personal data is used for this gate.
Source and evidence
Policy: restoration_value >= attention_cost. Browser source: aml-core/docs/aml-browser.js, blob e4e64a00d20718c1b74d8e5dd9f9dd08b1b84865. Read the before/after checkpoints and reproduction commands.
This is project-authored engineering evidence. The declared scores do not prove less cognitive load, better accessibility, more sales, or a better customer experience. A measured user study would be needed for those claims.