A decision audit, then a record that learns.
Every audit follows the same disciplined structure — on a term sheet or a care arrangement. Every audit you save makes the next one sharper. Here is exactly what happens.
Five dimensions of what you're missing.
⬡World Model Gaps
What your model of the situation leaves out — the facts you are treating as settled that were never checked, the variable you forgot was a variable.
▽Adversarial Risk
Who benefits from how you are being persuaded. The incentive behind the timeline, the introduction, the framing you did not choose.
◎Theory of Mind
What every other actor in the decision actually believes, needs, and is constrained by — modeled separately from what you wish they believed.
◫Decision state
Whether the decision is still yours, or already framed by someone else. Whose question are you actually answering.
◉Confidence vs Evidence
Where your stated confidence sits relative to the evidence on record. The gap between them is the assumption you have not yet named.
Every audit ends with the real question underneath the decision — the one you came in thinking you were answering, and the one you were actually avoiding.
The same findings, seen as structure.
Where DAUDIT stops being a tool and becomes a system.
One audit is useful. The compounding value is in what happens to every audit after you save it.
The Record
Capital and life, kept in separate drawers — each decision recorded with its stake, its audit, and when the time comes, its outcome. DAUDIT is the only thing that remembers to ask how it turned out.
ReadHow you decide
From your decisions, two profiles — capital and life, each built from its own record — of how you specifically decide: your confidence patterns, your recurring blind spots, what you update on and what you resist.
ReadOutcomes
A Brier score that measures your forecasts against reality. The honest answer to whether your decisions are actually getting better.
ReadYou've Been Here Before
When a new decision rhymes with an old mistake, DAUDIT says so — before you commit, not after.
ReadThe engine is not the difference
DAUDIT runs on Claude — the same class of model that agrees with you elsewhere. That is the point. A general assistant agrees with you not because of its model but because it was instructed to be helpful, and helpful almost always means agreeable.
This one was instructed differently: examine, never rule. Work only from what you actually said. Name the thing you have not weighed. Refuse the verdict even when you ask for it three times.
Same engine. Different orders. And because it is the same engine, you can check — take the same decision somewhere else and compare what comes back.