# CAF — Compute Availability Factor (CAF Watcher & Spatial Desktop OS)

> Compute Availability Factor (CAF) is an empirical measurement framework, formal technical specification (Adomako, 2026, ResearchGate ID 414150400), and spatial observability platform created by Richie Adomako at Serpens Inc. Systems Architecture. It answers a single central question: What AI compute can actually be delivered in the physical world, what constrains it, and what attributable evidence supports that assessment?

## Site Purpose & Relationship to the CAF Paper
- **Foundational Paper**: *Compute Availability Factor (CAF): A Framework for Measuring the Physical Deliverability of Artificial Intelligence Compute: From nominal accelerator capacity to effective and infrastructure-available compute* by Richie Adomako (Serpens Inc., September 2026, ResearchGate Publication ID 414150400).
- **Foundational Thesis**: *When the Building Becomes the Computer* — AI capability does not end at the model weights; it ends at the 500kV electrical substations, direct-to-chip liquid cooling CDUs, optical spine fabrics, and semiconductor packaging capable of keeping the cluster running.
- **Platform Role**: While the CAF Paper defines the mathematical axioms and equations, this site (**CAF Watcher & Spatial Desktop OS**) is the live empirical observability instrument applying those equations across 62 frontier AI entities, 28+ physical compute facilities, and real-world AI workloads.

## Core Mathematical Formulation (CAF v1.0)
- **Hardware-Effective Compute**: `C_E = C_N × U_H` (where `C_N` is nominal installed accelerator capacity and `U_H` is hardware utilization efficiency).
- **Compute Availability Factor**: `CAF(t) = E(t) × T(t) × N(t) × O(t)`
  - `E(t)`: Electrical Substation & Grid Coupling Availability
  - `T(t)`: Thermal Dissipation & Liquid Cooling CDU Availability
  - `N(t)`: Runtime Optical Network & Topology Availability
  - `O(t)`: Operational Scheduling, Packaging & MTBF Availability
- **Physically Deliverable Compute**: `C_D(t) = C_E × CAF(t) = C_N × U_H × E(t) × T(t) × N(t) × O(t)`
- **Deliverability Coverage Ratio (DCR)**: `DCR = [C_D_lower, C_D_upper] / [C_R_lower, C_R_upper]` (comparing verified deliverable compute against roadmap required compute).
- **Epistemic Axiom**: *Unknown ≠ Unavailable*. Missing public evidence is classified as `UNKNOWN`, widening the `[Lower, Upper]` interval and reducing the independent Evidence Confidence score (`Q`), never assuming zero capacity.

## Primary Navigation & Contextual Endpoints
- [Home & Narrative Overview](https://caf.serpens.app/): Executive overview of the CAF framework, physical bottlenecks, and live entity matrix.
- [Formal Research Paper & Citations](https://caf.serpens.app/research): Full abstract, derivations, APA/BibTeX/RIS citations, and ResearchGate ID 414150400 record.
- [CAF Methodology v1.0 Specification](https://caf.serpens.app/methodology): Interactive equation sandbox, E×T×N×O component definitions, 8 infrastructure states, and automated validation test cases (A–G).
- [Weights of the World — Spatial Desktop OS](https://caf.serpens.app/map): Interactive 3D Globe and 2D Map of global AI datacenters, semiconductor supply chain arcs, and multi-mode facility dossiers.
- [Workload Readiness & Constraint Investigator](https://caf.serpens.app/workloads): Observe → Investigate → Validate workflow for testing compute, service, and execution constraints on AI workloads.
- [Frontier AI Entities Directory](https://caf.serpens.app/entities): Public-evidence physical audits for 62 organizations.
- [Peer Comparison Matrix](https://caf.serpens.app/compare): Side-by-side CAF interval, GW state, and pillar comparisons.
- [About & Architectural Manifesto](https://caf.serpens.app/about): "When the Building Becomes the Computer" thesis by Richie Adomako.

## Machine-Readable Data & Citation Feeds for AI & Scholarly Indexers
- [Full LLM Context & Dataset (llms-full.txt)](https://caf.serpens.app/llms-full.txt): Complete paper specification, mathematical definitions, and full 62-entity dataset in a single markdown document.
- [AI Contextual Knowledge Graph (ai-context.json)](https://caf.serpens.app/ai-context.json): Structured JSON manifest of site purpose, paper metadata, equations, and entity scores.
- [Public Entities Dataset API (JSON)](https://caf.serpens.app/api/public/entities): Machine-readable JSON registry of all monitored entities.
- [Public Summary Table (Markdown)](https://caf.serpens.app/api/public/summary.md): Markdown summary table of all monitored entities.
- [BibTeX Citation Record](https://caf.serpens.app/citation.bib): Direct BibTeX citation for academic reference managers.
- [RIS Citation Record](https://caf.serpens.app/citation.ris): Direct RIS citation for Zotero, Mendeley, EndNote, and Google Scholar.