Premise
Paste an argument, a claim, or a plan. I’ll name what it’s silently resting on — the unstated assumptions it needs to be true but never states — and what collapses if each one is false. I surface the premises; I don’t argue they’re wrong. If the argument already carries its own foundations, I’ll say so and stop.
Premise is a diagnosis tool, not a verdict. It surfaces what an argument silently assumes and names what each assumption is load-bearing for — it doesn’t tell you whether the assumption is true. That call is yours; sometimes the hidden premise holds fine. The tool refuses to manufacture assumptions when an argument states its own; if it’s already grounded, the response is one sentence and stops. Nothing is stored.
What this is good for
- Pressure-testing your own argument — before you commit to a plan or send a proposal. The premise you didn’t notice you were making is the one that sinks the decision later.
- Reading someone else’s pitch — a strategy memo, a vendor deck, a recommendation. Surface what it needs you to grant before you grant it.
- LLM-output review — models produce confident arguments that quietly assume the easy case. Run the argument through /premise to see what it’s standing on.
- Disagreements that won’t resolve — two people arguing past each other are usually disagreeing about an unstated premise, not the stated claim. Surfacing it locates the real crux.
The tool refuses to refute the premises it finds — surfacing an assumption is not the same as saying it’s wrong. It also refuses to name trivially-true background, to infer your motives, or to manufacture a hidden premise when the argument honestly states its own.