Wayfinder — universal doubt layer
A second opinion for every action.
One call turns any text in 100+ languages into a calibrated verdict: act, review, escalate, block.
33 ms per decision100+ languages$0 self-hostedone forward pass
What it answers
Four policies cover nearly everything. Pick one to load it in the console.
Verdicts, fresh from the gate
“Billed twice for March. Refund it today or we cancel.”
Calibrated, not merely confident
Probabilities trained with proper scoring rules, then temperature-fit on held-out data. When the gate says 0.89, it means it. Most agents act first and explain later — the gate flips that order.
- Expected calibration error after fitting. Down from 0.466 raw.
- 0.081
- Languages clearing 3× random, routed automatically.
- 45/51
- A single decision on one GPU. Batched, about a millisecond each.
- 33ms
- Self-hosted weights, Apache-2.0. Safety stops being a line item.
- $0
From policy to verdict in three moves
- 01
Write the policy
A few typed questions in YAML: choice, score, or yes-or-no. No training, no prompts to babysit.
- 02
Make one call
Send any state in any language. The router picks the checkpoint; every question scores in a single forward pass.
- 03
Act on the verdict
High confidence moves on its own. Everything else lands in front of a human with the reason attached.
One container · one policy file · every stack
Ship calibrated doubt.
Wayfinder · doubt, as a service