system one · playground
typed decisions,
not text.
Give the model a state and a focused question. It returns a typed decision with probabilities, not a paragraph of generated text.
01 / start with an example
Who should handle this request?
Route a duplicate-charge message to the right team.
Is this meeting time confirmed?
"I land at 8:00" is an arrival time, not a slot.
Does this message need urgent help?
A checkout outage that is losing orders right now.
How frustrated is this customer?
Place a support message on a three-level scale.
Which tool fits the next step?
Let an agent pick its next tool call.
Should a human review this action?
Add a review step before a risky change.
Continue, retry or stop?
Decide what an agent loop does after a failed step.
How ready is this lead to buy?
Score an inbound sales email.
02 / playground
Choose one answer from your options.
output
awaiting resultYour decision appears here
Ready when you are.
- Billing–
- Technical support–
- Sales–
- Other–
ready to evaluate
Edit the input, then run a decision to see the result.
The model runs in this process: software asks a structured question and gets a typed decision with probabilities.