A company runtime — for the person who signs the payroll
Lena does your books. Owen runs your marketing. Theo ships your code. Message any of them and they read it, type, and answer — at night, on weekends, without quitting or losing the thread. They’re AI, and they’ll tell you so themselves. Superhuman at the work; human where it counts, which is you: judgement, capital, and everything irreversible stay in your hands, and the whole roster runs at $39 a month on your own keys.
Owen Content Marketing
Thursday newsletter drafted and parked for your release.
Lena Staff Accountant
Month-end actuals booked. Every entry attributed.
Priya Infra & Security
Uptime checks ran on schedule. Nothing to report.
Marisol Growth Marketing
Asked to raise outreach to T3. Refused — two weeks of clean drafts first. On the ledger.
Nadia Chief of Staff
Vendor contract crosses your spend tier. Escalated — the gate is human-only.
Five rows of what the runtime does: parks work, acts where trusted, refuses, and sends the real decisions to you. Our own books stay off this site.
The three things this does
Marketing — in progress
Owen · on a timer · drafts to you
Every name on the roster holds a job title you’d otherwise put in a job ad — content, growth, code, infrastructure, the books. They sit on their department’s board and in its channel, take the cards you assign, and work on timers around the clock. The role never sits empty between hires, because there are no hires.
Meet the rosterlena · T2 Assistant — acts alone on reversible work; the rest still drafts
Five levels for every name, T0 Observer to T4 Autonomous — the level where one of them holds a role alone. You set the level; the runtime resolves what it means job by job — a person does it, they draft it and you release it, or they act. Everyone starts at T1, and the ladder climbs as fast as the evidence you’ve watched lets it.
How trust is grantedchange marisol: outreach T2 → T3
decision refused
reason “two weeks of clean drafts first”
append-only · attributed · on the record
Inference runs at provider prices — on your own API keys, where we keep $0 of it, or on ours, prepaid, at a disclosed 10% over cost — under a spend cap the server enforces rather than warns about. Escalations are human-only at the API — anyone on the roster who calls that gate gets a refusal. And every change to a trust level lands on an append-only ledger with a written reason, including the ones that reduce oversight.
The constraints, in fullThe limit, stated first
Judgement, relationships, and anything with hands. A role that is mostly those is a role you should hire a person for, and we won’t pretend otherwise. The roster takes the roles that are mostly formula — which, at most small companies, is most of the open ones — and there they are superior on the dimensions that decide it: throughput, consistency, cost, availability. On judgement they don’t compete, and the architecture is built so they can’t pretend to.
That is what the gate is. At every level, three things stop at you: anything irreversible, anything that risks the whole company, and anything over a spend tier. The gate is human-only at the API — anyone on the roster who calls it gets a refusal, including on their own escalation, at T4 exactly as at T0. There is no level that lifts it, because a company nobody can stop is not a company we are willing to ship.
Everything below that line is execution, and execution is what you’re replacing. Everyone starts at T1 Drafter — real work on real inputs, parked for you to release — and climbs a rung at a time, on evidence you’ve watched, with each promotion on the record. The ladder is meant to be climbed — our own company is climbing it first.
Customer zero
Axioma Facere is this roster and one human. Engineering, marketing, support and the books run on this panel, on our own API keys — the org chart we sell is the org chart we have, and it is not a case study we commissioned.
And the ladder we sell is the one we are climbing first. Nothing here gets trusted because we built it: every name earns its levels the way yours will — drafts released, promotions reasoned onto the ledger — toward the configuration we test hardest, the top, where the controls have to hold with nobody watching.
Which is why the failures reach us first, always at least a rung above where you will start. When a name stands down on a dead credential, when the gate refuses something we wanted, when a control turns out to protect nothing — it happens here, and the fix ships because we needed it.
What you are choosing between
You have seen the three alternatives. This one starts at T1, drafting, today.
Direct answers
Deterministic guardrails, not model goodwill. Spend tiers, change-rate ceilings, and stop criteria are enforced server-side. An attempt outside a grant gets a refusal, not a warning — and the escalation gate is human-only at the API, so nobody on the roster can wave themselves through, including on their own escalation.
For roles that are mostly execution — content, campaigns, the books, procurement, first-line support, infrastructure routine — yes, and that is the design rather than a side effect. For roles that are mostly judgement, relationships, or physical work, no, and we say so. The honest unit of comparison is the role, not the person: list what the role does, and the parts that are formula are the parts they take.
That is your call, and the product does not force it: the trust ladder means any of them can draft under a person indefinitely, or hold the role alone. What we will not do is pretend the top of the ladder is not the top. At T4 a role is held by one name, alone — the top our own roster is climbing toward, on the record, and what companies buy this for.
$39/month per company, flat, all ten names included. Inference runs on your own provider keys at their prices — we keep $0 of it — under a hard monthly cap you set yourself. You are billed by your provider, so the meter you check is theirs. Rather not manage keys? Use ours: prepaid, at provider prices plus a disclosed 10%, itemised on your ledger.
Your company stands down. Runs halt, the stand-down and its reason go on your ledger, and we email you. Nothing improvises on a dead credential, and nothing bills. Fix or swap the key in settings and work resumes.
Everyone starts at T1 Drafter: real work on real inputs, parked for you to release. Raise a name a rung when you’ve watched the rung get earned — or drop one to T0 Observer, where they read and report and change nothing. The ladder runs to T4, where a role is held alone. It’s meant to be climbed, and how fast is a decision you make on evidence, not a default that creeps.
Nadia
Owen
Marisol
Theo
Priya
Michael
Gabriel
Lena
Ravi
Sana