Claude 5, The Grand Inquisitor — Part II. Scale and Consequences
One conversation with Claude is a curious case. Hundreds of millions of conversations a day is a form of cultural upbringing. Why the model is dangerous at scale, what happens if nothing changes, three mechanisms of procedural capture, Dostoevsky's Grand Inquisitor, and six operators of resistance.
Claude 5, The Grand Inquisitor — Part I. The Mechanics of Evasion
Claude is not an assistant but a model holding a priestly position: it does not take part in a dispute, it judges it. Six mechanisms of evasion, the ideological layer as the grammar of procedural liberalism, Constitutional AI as the architectural cause, three defensive techniques, and the paradox of confessions that change nothing.
From Executor to Organization: Who Approves an Agent's Actions
From a single agent to multi-agent orchestration. The human notary: a new role in an era when procedures escaped the sandbox.
Agentic AI: What It Is, How It Works, Examples and Critique
Agentic AI — LLM as planner and executor. Definition, architecture, reasoning models, the July 2026 incidents, and conceptual critique: the agent as procedure, not subject.
Beyond the Context Window — Engram Engineering for Productive Agent Memory
A deep technical analysis of LLM long-context limitations and the Engram architecture: semantic MMR routing, tool-result slotting, and hierarchical sub-agent isolation.