Simo glossary

Terms

• Judgment model: a model that answers typed questions about a situation with probabilities and values, instead of writing prose. Simo is one.
• System 1: fast, automatic intuition. In AI, a model that answers instantly with a verdict.
• System 2: slow, effortful deliberation. In AI, a reasoning model that writes step-by-step text before answering.
• System 1.5: Simo’s design: the System 1 reflex plus exactly the deliberation that acting requires.
• Calibrated probability: a probability that matches the real hit rate: at 0.8, right about 80% of the time. More.
• Calibration error: how far a model’s stated probabilities are from its real hit rate. Simo-1 Pro’s is 0.047 on Banking77.
• Choose one (Choice): pick from a predefined list; returns a probability per option.
• Yes / No (Boolean): a binary judgment with a probability for each side.
• Rate on a scale (Score): a rating on an ordinal or numeric scale.
• Fill a value (preview): a schema-constrained value the next action needs, such as a button label.
• Extract from the input (preview): values pulled from what is actually there, each with a calibrated self-check.
• Self-check: a calibrated score attached to a produced value, a second opinion in milliseconds.
• Simo-1: the default, real-time model (simo-1).
• Simo-1 Pro: the maximum-quality model (simo-1-pro).
• Act / ask threshold: the probability above which an action executes automatically, and the one below which a human is asked.
• Typed question: a question with a declared answer type, so the answer arrives in a shape code can use.