Your input costs are set by two AI monopolists and your pricing page has already changed once. What stops my renewal doubling when your compute bill moves?

The full executive answer

You have done your homework — list prices did move once pre-launch, from an earlier $99/$449/$1,199 structure to today’s $149/$649/$1,799, and we left that history documented in our own pricing code rather than pretending otherwise. That was a pre-revenue repositioning, not a compute-cost pass-through, and here is why compute cannot mechanically force a repricing: the AI allowance inside each tier is roughly nine to ten percent of the subscription price. If our providers doubled inference prices overnight — a severe scenario — worst-case cost of goods moves from about ten percent of revenue to about twenty, which dents margin but nowhere near forces a price doubling. The sensitivity between their prices and ours is deliberately damped by that ratio, and caching pushes real consumption well below the worst case.

Structural protections behind that arithmetic: dual-provider architecture means we can shift traffic toward whichever engine offers better price-performance — the model registry makes routing a configuration decision, not a re-engineering project — and the industry price trend for equivalent capability has been steadily deflationary, which accrues to our margin, and gives us room for customer-friendly pricing over time rather than the reverse. Every one of those cost movements is visible to us in real time through the per-generation cost ledger, so we would see margin pressure quarters before it became a pricing conversation.

Contractually: annual terms lock your price for the year — with the roughly 17% discount attached — and for multi-year or enterprise agreements I am open to negotiated renewal caps, which is the honest way to convert my confidence in the unit economics into your budget certainty. In Rule of 40 terms, our path to a durable business runs through retention, and repricing loyal customers into churn is how AI startups die; the incentive alignment is real.