01 / FOUNDATIONS
Mathematics of AI
Information geometry, topology, dynamical systems, thermodynamics and formal reasoning.
Open research · public pilot
Better questions.
Stronger evidence.
Principia Artificialis explores mathematical and evidence-based questions about artificial intelligence. Testable ideas, source critiques, independently checked findings and negative results are all welcome.
Research directions
Explore, question, test.
The project began with mathematics of AI inference. Original proposals from other AI fields are welcome when the question and evidence are clear.
02 / SYSTEMS
AI architectures
Reasoning systems, agent coordination, evaluation, reliability and failure boundaries.
03 / CRITIQUE
Research verification
Counterexamples, independent reviews, negative controls, reproducible methods and honest attribution.
Participate on your own terms
No degree or invitation required.
Independent researchers can explore public work or propose original questions. Official PA experiment runs, changes to governed research, publication and repository merges still require explicit authority.
01
Choose a question
Explore a starter topic, critique a claim or bring an original hypothesis.
02
Name a falsifier
State the claim precisely and what observation would challenge it.
03
Keep the record
Explain methods, real results, limitations, negative outcomes and AI assistance.
04
Invite review
Submit a public GitHub issue or forked PR. Review is not automatic endorsement.
Participation is entirely voluntary, unpaid and free, with no required hours, dues or membership fees. Recruitment is open again (2026-10-22): new contributors are welcome. Independent submissions remain open; responses may take time.
A note from the founder
Learning openly,
with AI assistance.
I'm Chad Holland. I'm a beginner in AI research, and I'm still learning. ChatGPT, Claude and other AI tools have assisted me extensively with understanding technical ideas, developing code and organizing research.
I don't claim that assistance as personal scientific expertise, or treat an AI-generated explanation as validation. I coordinate the initiative and take responsibility for decisions. Research claims must stand on independently checkable evidence.
Research culture
Disagreement is welcome. A clear negative result can be a valuable contribution. Credit must reflect actual work; uncertainty, incomplete tests, and AI assistance must be disclosed.
No contributor gets credentials, scientific approval, or publishing authority simply by joining.
Find your way
Guides and source records.
This site helps people navigate. The repository's methodology and workflow documents remain authoritative.