First — thank you for being here.
Sapien is an ambitious long-term research direction. It is not a finished product. It is an evolving idea that needs critical thinkers, researchers, builders, and honest skeptics to grow into something real.
Every contribution matters — whether it is a challenge to the core ideas, a refinement of the architecture, a formal proof, or a single sentence that makes the documentation clearer.
- Who Should Contribute
- Ways to Contribute
- What Sapien Needs Most
- How to Contribute
- Contribution Standards
- Opening an Issue
- Submitting a Pull Request
- Research Contributions
- Discussion and Debate
- Code of Conduct
Sapien welcomes contributions from anyone who thinks seriously about intelligence, learning, and AI architecture. You do not need a PhD. You do not need industry experience. You need curiosity and honesty.
Specifically valuable backgrounds include:
- AI and machine learning researchers
- Cognitive scientists and developmental psychologists
- Neuroscientists and philosophers of mind
- Knowledge graph and graph database engineers
- Symbolic AI and neuro-symbolic systems researchers
- Distributed systems engineers
- Education theorists and learning scientists
- Anyone who has built and trained language models and noticed something was missing
If you have challenged these ideas and found them wanting — your criticism is especially welcome. Sapien is strengthened more by honest disagreement than by agreement.
Theoretical contributions Formal specifications, mathematical proofs, theoretical arguments for or against components of the architecture. Papers, preprints, or structured writeups.
Architecture refinements Proposed changes, additions, or alternatives to the architecture documented in ARCHITECTURE.md. Must include reasoning — not just what to change but why.
Limitation analysis Deeper investigation of the known limitations listed in LIMITATIONS.md. New limitations identified. Proposed mitigations. Honest assessments of what cannot be fixed.
Implementation proposals Concrete proposals for how components of Sapien could be implemented. Prototypes, proof-of-concept code, benchmarks.
Documentation improvements Clarifications, corrections, better explanations. If something in the documentation confused you, improving it helps everyone who comes next.
Discussion and debate Opening issues to debate architectural decisions. Challenging assumptions. Proposing alternatives. Even disagreeing with the entire premise.
Current highest-priority research questions:
Grounding and embodiment How can Sapien acquire grounded meaning without physical embodiment? What role could simulated environments, multimodal perception, or robotics integration play? This is the deepest open problem. See LIMITATIONS.md §1 for detailed analysis.
Formal didactic episode specification A mathematically rigorous definition of the didactic episode learning loop. What are the formal termination conditions? How is epistemic closure formally defined and measured? See DIDACTIC_SPEC.md for the current proposal and its open problems.
Knowledge graph scalability How does the DAG scale to billions of nodes? What pruning, compression, and archival strategies preserve WHY chains while maintaining practical retrieval latency? See SCHEMA_SPEC.md §11 for current open problems in the schema.
WHY chain depth termination How do you define an axiomatic floor — the set of foundational concepts accepted without further justification? How is this floor selected and validated? See AXIOMATIC_FLOOR.md for the current proposal and its open problems.
Curiosity reward formalization A formal specification of the productive utility gate that distinguishes productive from pathological curiosity. How is semantic uniqueness measured? How is productive utility quantified? See REWARD_SPEC.md for the current specification and its open problems.
Contradiction resolution How does Sapien represent and reason about genuinely unresolved contradictions — contested science, philosophical disagreement, uncertain empirical claims? Probabilistic epistemology approaches are especially welcome. See CONTRADICTION_FRAMEWORK.md for the current framework and its open problems.
Generational drift measurement How do you measure reasoning drift across generations? What metrics indicate healthy evolution versus problematic divergence? See LIMITATIONS.md §3 for detailed analysis.
Step 1 — Read first Before contributing, read:
- README.md — project overview
- ARCHITECTURE.md — complete technical specification
- LIMITATIONS.md — known open problems
- DIDACTIC_SPEC.md — formal episode specification
- SCHEMA_SPEC.md — knowledge graph schema
- REWARD_SPEC.md — reward signal specification
- AXIOMATIC_FLOOR.md — axiomatic floor proposal
- CONTRADICTION_FRAMEWORK.md — contradiction framework
- Open issues — what is already being discussed
Step 2 — Find your contribution Check open issues for areas needing work. Or identify something not covered that you want to address.
Step 3 — Open an issue before large contributions For significant architecture changes, new research directions, or implementation proposals — open an issue first to discuss before investing time in a full pull request.
Step 4 — Fork and contribute Fork the repository, make your changes, and submit a pull request with a clear description of what you changed and why.
Honesty over agreement Do not contribute something you do not believe. If you think a part of the architecture is flawed, say so clearly. Honest criticism is more valuable than polite agreement.
Reasoning required Every significant change or proposal must include reasoning. Not just what you propose — why you propose it, what problem it solves, and what tradeoffs it introduces.
Acknowledge limitations If your proposed solution introduces new problems, state them. Sapien already maintains an honest limitations document. Contributions should maintain that standard of honesty.
Attribution If your contribution builds on published research, cite the work. If your contribution is an original idea, claim it. Credit matters.
Language English only for now. Clear, precise, and direct. Academic formality is not required. Clarity is.
Use issues for:
- Questions about the architecture
- Identified flaws or limitations not yet documented
- Proposals for architecture changes
- Discussion of open research problems
- Requests for clarification
Issue title format: [TYPE] Brief description
Types:
[FLAW]— identified architectural problem[PROPOSAL]— proposed change or addition[QUESTION]— request for clarification[RESEARCH]— relevant research to incorporate[DISCUSSION]— open-ended debate
Example: [FLAW] WHY chain termination has no formal stopping criterion [PROPOSAL] Probabilistic nodes for contradiction representation [RESEARCH] Schmidhuber 1991 — formal theory of creativity and curiosity
Pull request checklist:
- I have read ARCHITECTURE.md and LIMITATIONS.md
- My changes are consistent with the design philosophy in ARCHITECTURE.md Section 2
- I have included reasoning for every significant change
- I have updated LIMITATIONS.md if my change introduces new tradeoffs
- I have cited any external research I relied on
- My documentation changes are clear and accurate
Pull request description format: What this changes Clear description of what is being changed or added. Why Reasoning for the change. What problem does it solve? What limitation does it address? Tradeoffs introduced Honest assessment of new problems this change creates. References Any research, papers, or prior work this builds on.
Sapien is at the intersection of several active research fields. Relevant published research is especially welcome.
If you are aware of work that directly addresses an open problem
in Sapien — grounding, curiosity formalization, generational learning,
contradiction resolution, WHY chain depth — please open a
[RESEARCH] issue with:
- Full citation
- Brief summary of relevance to Sapien
- Which open problem or architectural component it addresses
- Your assessment of how it could be incorporated
Research contributions do not require a pull request. An issue is sufficient.
Sapien's adversarial collaboration principle applies to its own development. Disagreement is healthy. Debate improves the architecture.
The Forestritium Discord server has a dedicated channel for Sapien discussion:
For longer-form discussion, open a [DISCUSSION] issue on GitHub.
Sapien follows a simple code of conduct:
Be honest. Say what you actually think. Honest disagreement is more valuable than polite silence.
Be specific. Vague criticism helps nobody. If something is wrong, explain precisely what and why.
Be respectful. Challenge ideas, not people. The architecture can be wrong without the person who proposed it being wrong.
Credit others. If you build on someone else's idea, say so.
Acknowledge uncertainty. Nobody knows how to build human-like intelligence. Contributions that claim certainty where none exists are not helpful.
License: AGPLv3 Contributions welcome
Last updated: 29 May 2026
Author: Aarav | Forestritium