10B-model

People not parameters. Modeling the needs of everyone on earth.

This repository is the canonical, regeneratable home of the 10B macro model: a 258-country × 15-category satisfaction matrix that flows into a single human-needs embedding plus three downstream heads (unsupervised dynamics, supervised trajectory, personal-agent surfaces).

The committed data — every model.meta.json, model.run.json, runs/output.json, and forecast — lives under global/10B/. Read the data README for what the artefacts mean and how the pieces compose.

Repository layout

.
├── global/10B/          # The materialised model registry (data).
│   ├── countries/       # 258 country rollups + per-category cells
│   ├── categories/      # 15 category world rollups
│   ├── regions/         # 6 UN-geoscheme regional rollups
│   ├── needs-tree.json  # The schema model that ties categories
│   ├── runs/            # Top-level embedding + manifest outputs
│   └── ...              # See global/10B/README.md
└── scripts/10b/         # The Python pipeline that regenerates global/10B/

Regenerating the data

The pipeline is idempotent — re-running on the same input produces the same output, byte-for-byte. To regenerate:

# 1. Drop a CIA World Factbook countries dump at the repo root
#    (or point $COUNTRIES_JSON elsewhere).
cp /path/to/cia-factbook-countries.json ./countries.json

# 2. Run the full pipeline (writes back into ./global/10B/).
./scripts/10b/pipeline.sh

Each step is also runnable on its own:

Step	Script	What it produces
1	generate.py	Per-cell + per-rollup model.meta.json / model.run.json
2	scorer.py	runs/output.json for every cell + country rollups
3	extra_models.py	dynamics / supervised-trajectory / personal-agent models
4	dynamics.py	k-means cluster assignments + per-country peer lists
5	agent_context.py	Personal-agent connect/contribute/advocate surfaces
6	regions.py	UN-geoscheme regional rollups
	forecasts.py	Per-rollup forecast.json stubs
	manifest.py	Top-level runs/manifest.json summary
	validate.py	Schema validation across every meta + forecast file

Pipeline output target

By default, scripts write back into this checkout's global/10B/ directory. Override via environment variables:

• BWM_10B_ROOT — points at a checkout of this repo (the directory containing global/10B/). Use this when the script's location and the target tree differ — e.g. running pipeline-as-CI against a separate worktree.
• BWM_LOCAL_ROOT — points at a writable local registry root used by the bwm-server admin process. Setting this lets a re-run land directly in the writable fork without going through a manual import.
• COUNTRIES_JSON — absolute path to the CIA Factbook countries dump. Defaults to <repo-root>/countries.json.

Resolution order is BWM_10B_ROOT → BWM_LOCAL_ROOT → repo root.

Consumers

The data here is the source-of-truth for any downstream system that wants a complete view of the global 10B model. One known consumer is the business-world-models admin server, which mounts this repo as a read-only registry and lets users fork all-or-selected models into a writable local registry via its Browse Repos surface.

License

MIT.