Skip to content

Open research and design questions ​

PanSphaira combines business-process adaptation, reusable software, explicit execution controls and evidence about results. Its local proofs make selected parts inspectable; they do not settle the general research questions below.

Expressing requirements without losing meaning ​

How can a system translate a business need into a process variant while preserving constraints, ambiguities and competing interpretations? When should it request clarification instead of generating a plausible-looking plan?

Knowing when knowledge can be reused ​

Which sources, assumptions and counterexamples must accompany a candidate? How can changed conditions invalidate only the affected conclusions? What distinguishes a repeated pattern from an accidental success?

Dividing work between agents and deterministic software ​

Which tasks benefit from model-based interpretation, and which require fixed, repeatable logic? How should uncertainty cross that boundary without turning into an authorized action or a false success?

Separating common meaning from system-specific behavior ​

What can remain stable when a provider changes its fields, permissions or recovery semantics? How can a capability say precisely which bindings and combinations are supported?

Connecting analysis with improvement ​

How can observations suggest better processes while preserving the difference between analysis, proposal, approval and execution? Which outcomes would demonstrate an actual improvement rather than a persuasive explanation?

Evaluating the broader approach ​

A useful evaluation would compare a declared baseline and changed requirements, preserve failures and unknowns, test transfer to genuinely different contexts, and account for human effort and corrections. Test totals or active agents alone would not establish improved business outcomes.

These questions invite technical critique and collaboration. They are neither a claim of scientific novelty nor a promise that every topic is implemented.

Start with the overview, choose an example, and inspect the Canon, architecture and capability evidence. Questions and proposals can be raised in Discussions or through the contribution process.

Local synthetic proof of concept — not a production release or security certification.