A research program on intellectual plurality: why original thought is hard, why it has arguably become harder despite rising demographic diversity, and whether small insulated groups produce disproportionate originality.
originality/
├── CLAUDE.md Program orientation, ground rules
├── docs/program/
│ ├── desiderata.md Immutable program principles
│ ├── research_program_overview.md How the program was narrowed from a broader conversation
│ ├── originality_deepresearch_markdown.md Literature review + pathway DAGs
│ └── pathway_lineage_tables.md Reference tables: pathways → research lineages
├── whitespace_1/ LLM multi-agent simulation (future)
├── whitespace_2/ Empirical decomposition study (ACTIVE)
└── whitespace_3/ Theoretical reconciliation paper (future)
Each whitespace is an independent subproject with its own pyproject.toml,
Makefile, CLAUDE.md, docs/, tasks/, src/, and tests/. They do not
share a lockfile.
Whitespace 2 — empirical time-series study of ~5–10M OpenAlex CS papers (1970–2024): do demographic plurality, semantic plurality, and canonical concentration diverge over time? Target: 3–5 months, $50–500 compute.
cd whitespace_2
uv sync --extra dev
make testSee whitespace_2/CLAUDE.md for the study thesis, stage map, and current phase.
Studies are ordered by ascending cost/risk, with each prior study de-risking the next:
- Whitespace 2 (empirical, ~$500) — establishes whether the central decoupling claim holds in observable data.
- Whitespace 3 (theoretical, ~$0) — formalizes the tension between cumulative preservation and per-capita variance generation.
- Whitespace 1 (simulation, higher cost) — agent-based model of actuator homogenization, activated once the empirical and theoretical grounding is in place.
See CLAUDE.md (program) and docs/program/desiderata.md (immutable
principles). Each whitespace has its own CLAUDE.md and docs/desiderata.md
that extend these.