research-document
Research
Research
Status: Working draft
Research contains hypotheses, experiments, evidence, corpora, active questions, deferred insights, and failed hypotheses.
Research is exploratory.
Research artifacts should clearly state status:
- Hypothesis
- Under Investigation
- Provisionally Supported
- Foundation Candidate
- Superseded
- Failed
Operating system:
- Research Operating System
- Hypothesis Registry
- Research Operating Manual
- Research Constitution
- Research Dependency Graph
- Evidence Decision Records
- Research Queue
Product architecture:
Experiments:
- FE-011A LLM Blind Pilot
- FE-012A Blind Primitive Reasoning Extraction
- FE-012B Primitive Grammar Synthesis Experiment
- FE-012C Manual Multi-Model Replication Kit
- FE-012C Repeatability Run 001
- FE-013 Procedural Language Hypothesis
Evidence runs:
- ECR-000001 Multi-Layer Procedural Representation
- ECR-000001 Limited Review Workspace
- ECR-000002 Recognition Sensitivity
- ECR-000003 Representation Sensitivity
FE-013 manual execution:
- Use
run-instructions.mdandoperator-checklist.mdinresearch/experiments/FE-013-procedural-language-hypothesis/for manual console execution across GPT-5, Claude, and Gemini.
FE-012C reporting:
- Generated reports are written under
research/experiments/FE-012C-manual-replication/reports/.
FE-012C repeatability:
- Repeatability runs reuse the locked packet instrument and must not expose prior outputs to participating models.
- Dataset comparison outputs for repeatability run 001 are generated under
research/experiments/FE-012C-repeatability-run-001/comparison/generated/.
Validation:
- Framework Engineering Validation Program
- Adversarial Review Protocol
- Competing Theory Challenge
- Independent Replication Plan
- Claim Review Template
- Validation Roadmap
Theory: