Cascade answers may I take this — deterministically, in under a millisecond of decision compute, and it signs the answer. That is the entire product. Everything below is a consequence of keeping it that small.
Every one of these would make Cascade easier to demo and harder to trust. Determinism is only credible if there is nothing in the path that could vary.
Your planner decides what should happen. Your machines make it happen. Cascade is the gate in between, and it is the only part that has to be exactly right every single time.
Nothing above it changes. No migration, no new message bus, no rewrite of your autonomy stack. One call before you act.
Because every decision is deterministic and hashed as it is made, the audit trail is a by-product of how the layer works rather than a separate logging system that could disagree with it. That is why we can offer proof at all — and why nobody bolting a log onto a model can.
The free beta is live today: 1,000,000 lifetime signed decisions, no card, nothing ever auto-charges. The paid tiers below are published prices and launch soon — beta users get grandfathered pricing. The meter is the signed entry count in your own chain, so any future invoice is recomputable by you.
One signed decision per claim. A fleet of each claiming a minute is decisions a month.
Releases and holds are decisions too — one claim plus its release is 2 decisions. The free beta covers 1,000,000 lifetime signed decisions, no card. Paid tiers launch soon and pick up from there; beta users get grandfathered pricing.