Seats are a lie.

Seat pricing is a holdover from the era when the expensive thing about software was building it, and hosting was a rounding error. Back then, charging per login was a rough proxy for value, and nobody minded much. That era is over. What's expensive now is running the intelligence — the tokens, the compute — and that cost doesn't track how many people you invite. It tracks how much the AI does.

So seats lie in both directions at once. A one-person company sending a hundred emails a week costs me more to run than a ten-person company sending one. Seat pricing would overcharge the small one and undercharge the big one, and both would know it. That's not a pricing model. That's a thing you tolerate until someone offers you the honest version.

So there's one rate: twenty-five dollars per million units of work, trending toward fifty as the platform matures. A unit is a normalized measure of compute — mostly the AI's own tokens, plus the small change of storage and requests. No seats. No tiers. No feature gates. Everyone gets the whole platform on day one, and you pay for what the AI actually does.

One rate works because the distribution is lopsided in a way that makes anything fancier into accounting theater. For an active customer, the AI's own work is essentially the entire bill; everything else rounds to nothing. Per-category multipliers would add paragraphs of explanation without moving the number. The fewer variables on the receipt, the more you can trust it — and you can predict your bill with a calculator.

The meter is the part I'm proudest of. Every operation the AI runs is attributable: which generation, which page, which send consumed which units, and what that cost in dollars. When the AI works, the meter runs. When the AI stops, the meter stops. If you generate nothing this month, you pay nothing.

And you're not exposed to a runaway bill. You start capped at fifty dollars — a hard ceiling the AI can't spend past without you raising it. The AI can't outspend your trust. New organizations also get a starting credit to evaluate the platform with full access, so the first thing you spend is ours, not yours.

Honest about where this is: it's early alpha, and there's no billing during alpha — early access runs on a free sandbox while the metering and invoicing get finished. The rate itself is set. The meter already tracks every operation across the platform. What's still being built is the last mile of turning that into a monthly invoice, plus a couple of flat add-ons — single sign-on, data residency — that aren't consumption-shaped and are priced separately for the few who need them.

But the shape won't change, because the shape is the point. Pay for what the AI does. No seats. No tiers. No feature gates. The meter runs with the AI, and when the AI stops, the meter stops.