Follow a project
An online store, through the seven layers
A candle maker needs an online store that ships to the US and Canada. She has one conversation, and the seven layers do the rest of the work as one system.
The conversation
A candle maker tells Maestro, by voice, that she needs an online store that ships to the US and Canada. Maestro asks about her products, her prices and how she wants to be paid, and confirms the goal with her.
Layer by layer
| Layer | What it did for the store |
|---|---|
| L7 Maestro | Asked about products, prices and payments, and confirmed the goal |
| L6 Harness | Opened the project with one shared context for every model and agent |
| L5 ZB Model | Defined what a finished store must do, wrote the plan and routed each task |
| L4 Network of Intelligence | One model planned the checkout, another built the pages, another made the product images, and another tested every payment path |
| L3 Middle Layer | Connected the store to her payment provider and her shipping carrier, and recorded each task |
| L2 Security and Confidential Computing | Checked each action against the rules of her project, and protected her customers' details while the models worked with them |
| L1 ZB Compute | Ran the models and agents, and then the finished store |
Maestro told her the store was live.
The reference build
That request was the whole brief for Northwick Candle Co., our reference build.
| Tasks | 49 |
| Models | 14 |
| Agents | 6 |
ZB Model routed each task to the model or agent suited to it, and checked every result before the store went live.
After launch
The store learns from real use and starts its own improvements. When a check finds a case the store handles poorly, the same seven layers plan the fix, build it, check it and run it, and Maestro tells her what changed.