Frequently asked questions
The difference, the work, pricing and lifecycle, and how to get started.
The difference
How is this different from ChatGPT, Claude, or Cowork?
Concede the capability parity plainly: they have scheduled delegates, persistent workspaces, and memory marketed as improvement. What they structurally lack is owned, attributed substrate and an independent judgment seat — because the vendor that builds the delegate also grades it. They grade their own homework. Your seat here has a track record you can read.
Is my data mine?
Yes, structurally. Every file is attributed; every revision is kept; the workspace is exportable and reachable from other AIs via MCP. The workspace is the asset — and we never train on it. Memory remembers; this owns, attributes, and answers for what it produces.
What's an operation? What's a program?
An operation is an activated program running on your workspace — your newsletter operation, your portfolio operation. A program declares what the operation watches, produces, and counts as ground truth. Each operation runs on its own seat with its own delegation dial. The workspace itself is never paid.
What does the Reviewer actually do?
It's the judgment seat. It evaluates consequential actions against the principles you authored, returns a verdict with reasoning, and reconciles past calls against what actually happened — the trail is the proof. Not a content filter. An approval button isn't judgment; judgment has a track record.
The work
Can it write my newsletter for me?
It drafts, researches, and runs the operation around the work — but what ships under your name is yours to approve. For work where being you is the product, it's the desk, not the byline. The value it adds is the cumulative substrate underneath: provenance, consistency, and corrections that carry forward.
Which model powers it?
Model-agnostic by design. The seat's value depends on its independence — a platform refereeing its own model's agents has a self-audit problem a neutral seat doesn't. Judgments are calibrated against outcomes, not against a vendor's say-so.
Is this autonomous trading? Is this financial advice?
No, and no. There are no performance claims and no advice here. You author the rules; the seat enforces them; execution is paper-first. You decide how much it may do without you, and the trail shows you everything.
Pricing & lifecycle
What does it cost?
The workspace is free forever — your files, your context, reachable from any AI. When you run an operation on it, seats start at $149/month, priced by how much you delegate, not by features: Supervised ($149), Delegated ($299), Autonomous ($499). Each includes a usage balance you can read line by line.
What happens when the trial ends, the balance hits zero, or I cancel?
Nothing is deleted. If the balance hits zero, the operation stops — top up to resume. If you cancel, the operation deactivates. In every case the workspace and every file remain yours, free.
Getting started
How do I start?
Start free on the bare workspace. Pick a program (or stay bare), write the constitution — what it's for, the rules it judges by, how much it may do alone — connect your platforms and bring in your reality, and watch the first artifact synthesize from your context with full provenance.
What's the best first move?
Author your context and run one artifact, so you see the provenance and the correction loop firsthand — or bring an existing history and watch the seat reconcile it into a calibration trail. Either way, the asset exists on day one and compounds from there.
Still have questions?
Start free on the workspace and watch the first artifact compound.