Frameworks for systematic Human-AI Collaboration (HAIC) in complex domains. Each model is battle-tested, AI-native, and optimized for both human thinking and AI reasoning. Shared as my current best understanding - strong enough to use, incomplete enough to evolve.
Source Hierarchy Protocol: Trust structure for handling multiple information sources in AI collaboration. Establishes explicit precedence (user documents > user statements > retrieved content > training knowledge) to prevent silent contradictions and unmarked mixing. Works for both interactive conversation and reviewing existing documents.
Claim Verification Protocol: Calibrating AI confidence to reality by classifying claims by hallucination risk and auto-verifying high-risk categories. Addresses fabricated current events, wrong specifics, and behavioral patterns (sycophantic pivots, serial claims). Works for both interactive conversation and reviewing existing documents.
Prediction-Control Spectrum: A meta-control loop for balancing model-driven planning vs. feedback-driven iteration. Helps you locate where you are between prediction and control modes, and adjust based on context. Applications to software engineering and innovation.
Expedition Protocol: Context management for long-running human-AI collaboration under uncertainty. Provides structure through four living documents (state, decisions, sessions, guide) and disciplined context budgeting. Solves context loss, drift, and decision archaeology in multi-session AI work.
More frameworks coming as they're ready.
Learning in public: My journey into AI-augmented work and multi-domain integration.
Finding collaborators: Looking for people working on systematic human-AI collaboration protocols.
Inviting criticism: These frameworks improve through exposure and scrutiny.
These are thinking tools, not prescriptions. They help you:
- Surface hidden assumptions
- Locate yourself when stuck
- Navigate complexity through feedback loops
Test them against your reality. Adapt ruthlessly and discard what doesn't work.
These documents work as reasoning context:
- System prompts for specialized agents
- Reference material for complex decisions
- Templates for protocol development
Each framework has clear problem statements, diagnostic patterns, and correction strategies.
Every framework here has been:
- Created through human-AI collaboration
- Tested in real practice (my work across software, investing, photography, AI)
- Refined based on actual use
This is not AI slop. Each piece represents genuine intellectual work - ideas wrestled with, tested, and distilled through personal use.
AI's role: Exploring concept space, testing consistency, refining language. Human's role: Problem identification, core insights, judgment, curation.
We live in a world of irreducible complexity. You cannot predict your way through it, but you can navigate it through tight feedback loops and systematic sense-making.
These frameworks favor:
- Sensing over predicting (when possible)
- Navigation over optimization (in uncertain terrain)
- Protocols over intuition (for reproducible collaboration)
On AI: Not a replacement for thinking, but an integration engine that enables sustained work across multiple domains through systematic protocols.
Interested in:
- Criticism and counter-examples
- Related work I'm missing
- Collaboration on similar problems
Open an issue to discuss.
Use these frameworks in your work, adapt them, then share them with attribution. They're most valuable when widely distributed and tested.
Software engineer with math background, working on integrated life across multiple domains. This repository is part of that experiment - developing systematic approaches to multi-domain work under uncertainty.
Frameworks will be added, refined and sometimes removed as learning continues.