Foundations
System 1.5: the reflex, plus exactly what acting needs
Psychology splits thinking in two. System 1 is fast intuition: instant, automatic, always on. System 2 is slow deliberation: effortful, step by step, expensive. Today’s AI gives software only the extremes. Simo sits between them.
The two extremes
System 1 models answer instantly with a probability, but only ever a verdict: yes, option B, level 3. This is the approach Jev made famous for decision-making. Fast, but there is nothing to act with. Which button? How far to scroll?
System 2 models, the reasoning giants, can do anything, but write a few thousand words to themselves before every answer. That takes seconds and dollars for what is usually a one-word decision, and “I’m confident” is not a number.
What System 1.5 means
Simo is the reflex plus exactly as much deliberation as acting requires.
- It judges like System 1: one pass over the situation, calibrated probabilities, milliseconds.
- It acts like a sliver of System 2: the chosen action comes with its typed arguments, extracted facts come straight from the evidence, and every produced value can carry its own calibrated self-check, a second opinion still in milliseconds.
- It never pays the System 2 tax: no chain of thought, no essays, no seconds of silence.
The same question, three ways
| System 1 | Simo (System 1.5) | System 2 | |
|---|---|---|---|
| Asked | What should the agent do next? | What should the agent do next? | What should the agent do next? |
| Returns | option B | press: 0.89, scroll: 0.06, stop: 0.05 and button: "Place order" | A few thousand words, then “Press the Place order button.” |
| Speed | Instant | Milliseconds | Seconds |
| Typed arguments | No | Yes (preview) | By parsing the prose |
| Calibrated number | Verdict only | Every decision | No |
Why it matters for software that acts
Software that acts is full of judgment calls: is the task finished, which team handles this, is this action safe, which button, what is the order ID. A reasoning model asked “what should the agent do next?” writes a few thousand words and then tells you. Simo answers directly, and tells you how sure it is, with a number you can build a system on.
That also makes Simo the natural fast system in front of a slow one. See why we built Simo and the Simo vs reasoning models comparison.
Frequently asked questions
What is System 1.5 in AI?
A model design that judges in one fast pass like System 1, and adds only the deliberation acting needs (typed arguments, extraction, self-checks) without System 2’s step-by-step reasoning text.
What are System 1 and System 2 thinking?
System 1 is fast, automatic intuition. System 2 is slow, effortful deliberation. The terms come from psychology.
Is Simo a reasoning model?
No. Simo does not write chain-of-thought. It answers each question directly with a calibrated probability.