Vendor Capability Matrices
Loop Engineering by Fabio Correa — the Applied AI foundation volume on working with AI through framing, feedback, verification, memory, and human judgment.
Compare current reasoning-model, agent, computer-use, and MCP surfaces here. Each matrix carries a verified date so readers can judge freshness; the catalog is reviewed quarterly.
Reasoning Models
OpenAI o-series, DeepSeek R-series, Anthropic extended thinking, Gemini Flash Thinking, GPT-5 reasoning modes. Capability deltas across visible-vs-hidden chain, thinking-budget controls, cost variance, and Anthropic's faithfulness caveat from Claude's extended thinking.
Open Reasoning Matrix →Agents and Computer Use
Claude Computer Use, OpenAI Operator/CUA, Project Mariner, Devin, plus the open-source agent frameworks. Capability deltas across confirmation-gate patterns, monitor-model implementations, sandboxing, and side-effect controls.
Open Agents Matrix →MCP Adopters
Cross-reference to the MCP server catalog. Block, Apollo, Zed, Replit, Codeium, Sourcegraph, plus community projects. Which capabilities each implements, which client surfaces consume them.
Open MCP Catalog →Loop Orchestration and Control
Market vocabulary now separates orchestration frameworks, durable execution engines, and governance layers. LangGraph, Temporal, CrewAI, Inngest, and control-plane tools answer different questions: how the loop moves, how it resumes, and what each step is allowed to do.
Map the Terms →The reasoning-models and agents matrices are live; the MCP-adopters cross-reference points into the dedicated MCP catalog. Refreshed quarterly; corrections welcome via errata.