MCP Is the Layer Where AI Stops Being a Chat
After running an 83-tool Model Context Protocol server in production, the model is no longer the product. The tools it can reach are. Most organizations are still treating AI as a chat box.
Research & Analysis
Analytical reports, operational frameworks, and research notes from WBA's independent practice.
When Agent Skills Lose to Simpler Guidance
A local Sigil benchmark found that, for summary and validation work, a small always-on guidance layer plus MCP outperformed manual invocation of a dedicated agent skill.
Multi-Model AI Memory: What the Implementation Record Reveals
Most organizations deploying multiple AI models discover that each model operates in isolation — context evaporates between sessions, decisions are repeated, and institutional knowledge fails to accumulate. This analysis examines why multi-model memory is an organizational infrastructure problem and what the implementation record reveals about durable solutions.
From Epistemic Debt to Epistemic Instrumentation: MCP as a Business Pattern
Organizations often keep sounding confident long after they stop knowing. MCP — implemented with safety tiers, structured errors, and audit-grade logging — can be a practical step toward epistemic instrumentation.