Exact
The computation is deterministic: the same declared world and analytical contract produce the same result, with the run identity recorded.
Why Kindynos
The hardest error to catch is the one that sounds right.
I produced an AI-assisted report that appeared sound, only to discover later that parts of it were fabricated. That experience exposed the problem Kindynos is designed to address: analytical output can look polished, rigorous and authoritative without having a defensible path from evidence to conclusion.
The problem is not simply that AI can be wrong. Any analytical method can produce errors. The deeper problem is that a supported result and an invented one can look equally authoritative.
AI has improved. The risk remains, and greater fluency can make fabrication harder to detect.
The requirement
Where a result is relied on, it must survive examination.
Analysis that supports a regulated or high-stakes decision must be reliable, not merely persuasive. It must be defensible, independently challenged, and traceable from evidence to figure.
Kindynos was built around that requirement — a derivation someone else can follow.
Defensible
Every figure comes from a named, versioned model with declared inputs and coverage.
Auditable
Inputs, assumptions, model lineage and evidence remain attached to the result, so an examiner can trace it rather than accept it.
The idea
What is the answer allowed to depend on?
Controlled Conditionality separates the conditions that belong to the problem from those that belong only to the machinery running it.
Problem variables
They are supposed to change the answer.
Assets, obligations, stakeholder relationships, evidence, events, conditions, assumptions and models all carry analytical meaning.
Incidental variables
They should not change the answer by accident.
A controlled architecture eliminates request composition, ordering and execution environment as dependencies, fixes them, or declares them explicitly.
The boundary
A language model may
read · resolve · explain
It never
produces the value
Every analytical value comes from an explicit model against declared state — never from a language model.
No figure originates in a language model; no explanation becomes part of the analytical basis.
Interpret
Read the world.
Extract observations, obligations, relationships and candidate causal paths from complex evidence.
Reason
Resolve semantic questions.
Where language, ambiguity and contextual judgment genuinely require it.
Compute
Call explicit models.
Deterministic numerical, contractual or accounting transformations, where the transformation itself must be auditable.
Explain
Communicate the result.
Translate governed computations and evidence into useful explanations without making the explanation itself the basis of the analysis.