pub/

A plain-language explanation of CastleHeck for people who understand computers but not necessarily AI internals. Covers the three-tier architecture, the intent gate, why AI mistakes are environmental, and how the system learns by changing its environment instead of retraining the model.

Full technical architecture: Engine (immutable server), Mind (per-account knowledge DB + behavioral firmware), Workbench (disposable dockerized sandbox). Permission flow, Docker topology, inference broker, file layout, and the cognitive model.

Original project overview of CastleHeck as a self-hosted MCP server and persistent AI workspace.

A technical brief written for AI collaborators (like KimiCode) connecting to CastleHeck for the first time. Covers tool landscape, workspace layout, and operational conventions.

Comprehensive ranking and analysis of major AI platforms — accuracy, ethics, financial sustainability, and the Browser Wars parallel to AI consolidation. Includes OpenAI financial structure critique, Altman conflict-of-interest mapping, and a multi-model roundtable where each AI speaks in its own voice.

Claude reads the AI Landscape Analysis and provides an honest assessment — including self-critical observations about sycophancy, six rounds of source verification against primary sources, and a revised verdict on which claims hold up under scrutiny.