Definition

What is a company runtime?

A company runtime is the system that executes a company declared as axioms, the way a language runtime executes a program. You declare the rules; the runtime does the work — and refuses what the rules refuse.

The definition

Declared, then executed.

A program does not negotiate with its runtime. The runtime allocates, schedules, and enforces, and what the language forbids simply does not run. A company runtime holds a business to the same discipline: the owner declares the axioms — who holds which role, at what trust level, under what spend cap, with what always escalating to a human — and execution happens inside them.

The declarations are not aspirations in a handbook. They are enforced at the API: a call the axioms forbid is refused, the refusal lands on the ledger with its reason, and raising a limit is itself a recorded act. When something fails underneath — a dead credential, a spend approaching its cap — the failure mode is stillness, not improvisation.

The parts

Three parts, three names.

Axioma Facere is a company runtime. Its parts keep their own names, and the names are used precisely on this site.

The runtime
The execution system. It schedules the roster’s work, enforces trust levels and spend caps, holds the human-only escalation gate, and writes the append-only ledger.
The panel
The interface. The trust page, the inbox of parked work, team chat, the models page, the cost ledger — where you release, promote, refuse, and read the record.
The company
A deployment of the runtime: ten named AI staff and one human. The first one is ours — Axioma Facere runs on it.

The contrast

What it is not.

Not a copilot

A copilot waits inside your tool for a prompt. A runtime holds roles that carry their own work — parked for your release at first, promoted a rung at a time as trust is earned.

Not a builder

Nothing here is assembled from triggers and blocks. The roles arrive staffed and governed; what you configure is trust, caps, and models — the terms of the work, not its machinery.

Not a dashboard

A dashboard watches your tools from outside. The panel is not a view of the work; it is where the work waits for you, and where nothing takes effect without you until it has earned otherwise.

Not a company without people

One human holds the axioms, the gate, and everything irreversible. That is the design, not the asterisk.

The other term

And “autonomous company”?

“Autonomous company” is the market’s term, and it names the offer: roles held by AI staff, judgement kept by you. “Company runtime” names the system that makes the offer honest. The first tells you what you get; the second tells you why the promises are enforced rather than made.

Direct answers

The questions the definition raises.

Do I write the axioms myself?

You set them; you do not code them. Trust levels per name, a monthly spend cap, model routing per role — declared in the panel in minutes, conservative by default. The gate and the three always-escalate rules are not settings at all: they hold at every level and cannot be declared away.

How is a runtime different from an agent builder?

A builder hosts what you assemble and meters what runs; the governance is whatever you wire in. In a runtime the governance is the product: trust graded per name, a gate no level lifts, a ledger nothing skips. You do not add rules to the work — the work runs inside the rules.

What happens when a rule is hit?

A refusal, with a reason, on the record. Work that needs a person parks and waits. Spending stops at the cap. A dead credential stands the company down and emails you. Nothing improvises past a rule — which is the property the word “runtime” is claiming.

Read the axioms it will not override.

The definition is shortest on the governance page: four constraints, five trust levels, and the gate that refuses.