Systems that can prove
their reasoning
Three platforms, built on one conviction: a decision that cannot be examined should not be trusted, however confident the system making it sounds.
Heuristic Systems builds software for judgements that are normally asserted and rarely evidenced, whether a student has learned, whether a claim is supported, whether work is worth doing next. In each case the product is not the answer. It is the ability to show how the answer was reached, to someone who doubts it.
Three platforms
Different domains, the same discipline: define the criteria before gathering evidence, keep the evidence attributable, and leave a person accountable for the conclusion.
Heuristic Learning Systems
apto “I adapt”
Adaptive learning that can show a student reached a capability, rather than reporting that they reached the end.
Six non-compensatory gates. Every grading run reproducible from its logged hashes.
ExploreHeuristic Research Systems
probo “I prove”
An agentic research methodology that keeps evidence separated by origin, so a conclusion can be traced to what produced it.
Sources scored independently and published side by side, never blended.
ExploreHeuristic Planning Systems
pono “I set”
Programme management for AI-driven development, where agents are first-class actors and completion requires evidence.
Priority is a calculation from cost of delay, not the loudest request.
ExploreWe are our own first customer
Our delivery runs on pono, including the build of the other two platforms. Work is prioritised by the same economic calculation we describe to customers, and nothing is marked complete without the evidence to support it.
That arrangement is useful mainly because it is unforgiving. A governance rule that is tedious in practice gets discovered quickly when you are the one obeying it.
How that worksGrounded in examined research
The learning platform rests on doctoral research and its supporting corpus: a body of work tracing how education came to measure completion instead of capability, and what a system owes a learner once AI mediates the relationship.
We publish that corpus as it is released, including the parts still awaiting formal publication, marked as such.
Read the research