We build Nova AI systems

The two-layer architecture

Every Nova system has exactly two layers. They are always separate and never merged.

• Layer 1 — Nova (technical, silent)

  The working layer: build bots, maintenance bots and autonomous tasks. It never talks to people. Once it's started on a job, it builds the thing, then keeps checking its earlier builds, adds new technology and repairs what breaks, without being asked again.

• Layer 2 — Nova AI (persona, voiced)

  The face of the line: a named, voiced persona that speaks with people. It takes requests and triggers the technical layer, but it does not do the technical work itself.

Human speaks to the persona. The persona triggers the technical layer. The technical layer executes, then keeps evolving and repairing everything it has built.

Why split it in two

• Safety. The part that talks can't quietly rewrite the system, and the part that builds can't talk anyone into anything.
• Clarity. People deal with one consistent personality, however many bots are working behind it.
• Self-repair. One action causes a reaction that serves evolution, repair or maintenance. The system keeps itself healthy.
• Low cost. The technical layer runs on free and low-cost model APIs and free hosting. Paid capacity can be switched on later without a rebuild.

What a Nova build includes

• Nova technical layer

  Build and maintenance bots set up for your project, with scheduled self-checks and repair passes.

• Nova AI persona

  A named persona with its own voice that speaks for your system, on your site or your channels.

• Model routing

  Each task goes to the right free model: reasoning, code, fast inference or fallback.

• Plain-HTML interface

  A front end that any person, search engine or AI can read.

The same pattern runs inside the Federation: Agent Zero is a silent middleware layer, and its voiced persona is Spine. Read more on the Process page.