Skip to content

Platform overview

The seven layer Intelligence Stack takes an AI project from a conversation to a running solution.

One layer listens, three layers of intelligence do the work, and three layers of infrastructure connect it, protect it and power it.

An empty orchestra stage in warm light, the instruments resting and a conductor's podium at the centre.

The conversation

The stack starts with a conversation, on the web, in a mobile app or on your phone.

You describe your project once, in your own words. Your account is your phone number, so the project follows you from a call to the app.

Conversation
  1. youMaestroyour ideareportsgoalyouMaestroyour ideareportsgoal
    L7

    Voice AI Work Assistant Maestro

    Understands what you want to create

Intelligence
  1. one contextbuildswritestestsa Harness of its ownown modelsown dataown rulesown language
    L6

    Multi model, multi agent Harness

    Runs the project, for a person, a company or a nation

  2. plansrouteschecksassembleslearns
    L5

    ZB Model

    Plans the work, routes each task and checks each result

  3. taskresultReasoningCodingCreating imagesTesting software
    L4

    Network of Intelligence

    Does the work

Infrastructure
  1. modelagentone way inmemorytoolsdatacompute
    L3

    Middle Layer

    Connects every model and agent to memory, tools, data and compute

  2. actionpolicyapproved actionsprotected areadata
    L2

    Security and Confidential Computing

    Sets what the AI can access and do, and protects data while it is in use

  3. 64 GPUsopen modelsagentsfinished product
    L1

    ZB Compute

    Runs everything above it

ZB Model

ZB Model is the intelligence that plans the work, routes each task and checks each result.

It turns your goal into a plan, splits the plan into tasks, decides what a good result looks like for each one, and checks every result before it moves on.

The Network of Intelligence

The Network of Intelligence brings models and agents from every major AI maker to your project.

ZB Model routes each task to the model that does it best. The strongest models take the hardest tasks, fast, low-cost models take the everyday ones, and all of them share one memory of your project.

ZB COMPUTETHE COMPUTE INFRASTRUCTURE ONE JOB Maestro THE ENTRY POINT AI MODELS AND AI AGENTS

Security and Confidential Computing

Every action follows your policy, and your data stays protected while the AI works on it.

You decide who has access, which models and data a task uses, and how much the work spends.

ZB Compute

ZB Compute runs the work on any cloud, in your private cloud or inside your border.

It is the compute infrastructure for every model and agent. It runs open models and gives every agent its own sealed computer.

Where a job runs

Improvement

The stack gets better with every project it runs.

Your project improves from real use, and ZB Model learns from each finished project how to plan the next one.

The ZB Stack platform

Your project starts with one conversation.

Tell Maestro what you need on the web, in the app or on your phone, and it takes your project from the first question to a running solution.

Talk to Maestro
A conductor's baton resting on a closed score beside an inkwell, by a window at sunset