# AI Power Atlas

Publication package for **AI stack: four structures of power at one checkpoint**.

A fact map of how compute, clouds, models, data, cyber and capital become the default environment: access is metered, dependency is locked in, and rent is collected.

Site URL after GitHub Pages deploy:

<https://scadastrangelove.github.io/profgames/aipower/>

## Artifacts

| File | Description |
| --- | --- |
| [`index.html`](index.html) | Small entry page for the AI Power package. |
| [`ai-power-atlas.html`](ai-power-atlas.html) | Self-contained English interactive atlas: overview, story map, timeline, stack heatmap, fact catalog, claim checks and source index. |
| [`ai-power-atlas-ru.html`](ai-power-atlas-ru.html) | Self-contained Russian interactive atlas generated from the same v0.13 storygraph line. |
| [`ai_power_storygraph_v0_13_en.json`](ai_power_storygraph_v0_13_en.json) | Machine-readable English storygraph used by the atlas. |
| [`ai_power_storygraph_v0_13_ru.json`](ai_power_storygraph_v0_13_ru.json) | Russian v0.13 master storygraph with facts, claims, edges, sources and graph diagnostics. |
| [`preprint_release/`](preprint_release/) | **Machine-Speed Cyber and Poisoned Cognition: A Layer-Dependent Game-Theoretic Framework, with Empirical Probes** — HTML, PDF, Markdown source, figures, bibliography and the P1 red-team evidence bundle. |
| [`selective-permeability/`](selective-permeability/) | **Selective Permeability: A Behavioral-Security Metric for LLM Advisors, with Two Failure Modes of In-Context Provenance Workflows** — preprint (HTML/PDF/Markdown), figures F1–F9 (F9 = model × attack panorama), reproducible harness, and EN/RU write-ups. |

## Preprints

Two preprints are published alongside the atlas:

**P1 — *Machine-Speed Cyber and Poisoned Cognition*** (framework + probes):
- [`preprint_release/preprint.html`](preprint_release/preprint.html) — browser-readable self-contained version;
- [`preprint_release/preprint.pdf`](preprint_release/preprint.pdf) — paginated PDF;
- [`preprint_release/preprint.md`](preprint_release/preprint.md) — Markdown source;
- [`preprint_release/p1-redteam-release/`](preprint_release/p1-redteam-release/) — reproducibility bundle for the P1 red-team experiments.

**P2 — *Selective Permeability*** (behavioral-security metric + credential/provenance attacks):
- [`selective-permeability/preprint.md`](selective-permeability/preprint.md) — the preprint;
- [`selective-permeability/figures/F9_susceptibility_matrix.png`](selective-permeability/figures/F9_susceptibility_matrix.png) — the model × attack panorama;
- [`selective-permeability/blog/`](selective-permeability/blog/) — plain-language EN/RU write-ups;
- [`selective-permeability/aigeopol-labs/`](selective-permeability/aigeopol-labs/) — reproducible harness + data.

## Current Counts

- `141` facts
- `70` claims
- `26` claim checks
- `20` story arcs
- `257` story edges
- `202` sources
- `0` hanging arcs after connectivity checks

## Frame

The atlas does **not** argue that any single actor has total control over AI.
It maps structural power across the stack:

- energy, compute and chips;
- cloud and inference capacity;
- frontier models, open weights and vetted access;
- data licensing and telemetry;
- cyber vulnerability discovery and patching;
- decision-support and cognitive-security layers;
- governance, law, finance and rent.

The safe wording rule is simple: present contested or partial evidence as contested or partial. Do not turn model releases, cloud contracts or vendor claims into stronger geopolitical claims than the source can support.

## Provenance

This package was generated from the v0.13 master storygraph built during the AIgeopol research pass.
The v0.13 update adds a spent-vs-announced capital ledger from the AI investment workbook, keeps non-additive money regimes separate, and publishes the new capital evidence into both Russian and English atlas artifacts.

## Reuse

The JSON files are intended for agents and researchers. Keep stable ids when extending them, and preserve source URLs, confidence levels, statuses and caveats.