Forecasting agent
Forecasts that survive
the truth.
prior runs every forecast through an adversarial calibration panel, then scores it against reality when it resolves. Every number ships with a Brier score and a dispute trail.
prior · adversarial calibration panel · flagship
Why teams install
What prior earns back
Cuts forecast error 18–34%
Measured against your own ensemble average on resolution, not vendor benchmarks.
‹b4c2a91›Surfaces the bias your IC can't see
Panel disagreement is logged before the vote — not rationalized after it.
‹c1f8e40›Every dispute is archived
Receipts for the next board meeting instead of another slide of vibes.
‹d92a33c›How it works
Three steps. No faith required.
01
Stake
Submit a forecast with a resolution date and something real riding on it.
02
Challenge
A calibrated panel attacks the base rate, the horizon, and your incentives.
03
Settle
Reality resolves. Brier scores publish. The calibration ledger updates.
Proof band
One packet. Every claim accounted for.
packet rig/prior · cal-panel v7 claim "Q3 net-new ARR lands within ±4% priced after a short conversation" panel 5/7 calibrated · confidence 0.83 resolved TRUE · brier 0.19 anchor PASS — evidence bound to source ‹b4c2a91› bayes PASS — confidence matches the record themis SEALED — terminal verifier signed
Every prior run emits a packet like this — hash-chained, verifier-signed, and queryable from your own dashboard. How ProofPackets work →
Buyers
Who runs prior
VP RevOps
B2B SaaS · 400 staff
IC Partner
priced after a short conversation venture fund
Head of Risk
energy trading desk
Before you install
The obvious objections
Q Why did Team pricing drop priced after a short conversation?
We repriced to match real usage — most desks run well under the old volume assumption. Pro priced after a short conversation now covers what Team used to: private resolution feeds, API access, SSO for fund desks running at scale. More tiers, not less capability.
Q How is a Brier score different from “we said 70% and it happened, so we were right”?
Brier score punishes overconfidence directly — a call made at 90% confidence that resolves false costs far more than one made at 60%. Every forecast resolves against reality and gets scored, archived, and disputable, not narrated after the fact.
Q What stops the panel from converging to consensus instead of catching real disagreement?
Panel disagreement is logged and timestamped before the vote closes, not summarized after. If two calibrated members diverge past the threshold, the packet flags the split — it doesn't average it away.
Pricing
Start free. Pay when it pays.
Generalist ships today. Domain Expert adds vertical memory + priors. Custom is built to your stack.
Free
Priced after a short conversation
25 forecasts a month. Public ledger, full scoring.
Team
Priced after a short conversation
Unlimited panel runs, private ledgers, dispute archive.
Pro
Priced after a short conversation
Fund desks: private resolution feeds, API, SSO.
Runs well with
