Use cases
Moderation and policy screening
Thousands of items an hour, auditable probabilities, and no essays to parse.
How it works
For every item, ask one Yes/No per policy: does this violate it? Simo reads the item once and returns a calibrated probability for each policy.
P(violates policy) for every policy → one probability eachThresholds you can defend
- Above the threshold: auto-action.
- Near the line: the review queue, with the probabilities attached.
- Below: pass.
Because the probabilities are calibrated, you can audit the decision and move the line deliberately. See calibrated probabilities.
Which model
Simo-1 for the stream. Use Simo-1 Pro for hard cases near the line.
Frequently asked questions
Does Simo write a reason for each decision?
No. Simo never writes prose. It returns a calibrated probability per policy, which you store and audit.