AAIF Ambassador Cohort 2026
About This Publication
thinkinginsidethebox.ai is a technical publication on agentic AI, hosted by Vincent Caldeira as part of the Agentic AI Foundation Ambassador program and open to collaborating authors. The site covers enterprise agent architecture, open protocols, and the engineering practices needed to run autonomous systems safely in production.
Editorial Philosophy
In consumer design, "thinking outside the box" is praised. In regulated enterprise environments — especially financial services — letting autonomous agents run unconstrained is an operational and security risk. For agentic AI to be trusted in production, we must design systems where agents reason, plan, and execute inside strict, deterministic, and inspectable boxes.
Topics
This publication focuses on the following editorial themes:
-
Agentic Safety
Runtime safety, policy enforcement, and governance patterns for production agentic systems.
-
Metacognition
Self-correcting cognitive loops and reflective reasoning inside bounded permissions.
-
Evaluation Driven Development
Testable, measurable agent behavior validated through structured evaluation harnesses.
-
Spec-Driven Development
Phased specification-to-implementation workflows where agents generate code from explicit, governed specs.
-
MCP 2026-07-28
The July 2026 Model Context Protocol revision — stateless core, Multi Round-Trip Requests, header-based routing, and enterprise-scale agent tooling.
AAIF & Open-Source Projects
Articles frequently reference and contribute to these open projects:
-
MCP
Model Context Protocol — interoperable tool and context integration for agents.
-
agentgateway
Architectural control plane for intercepting and governing agent tool execution.
-
AGENTS.md
AAIF standard for universal agent context — stack, commands, boundaries, and persona registries.
-
GitHub Spec-Kit
Spec-driven development toolkit — phased specify, plan, and implement workflows for AI agents.
-
OGX
Open GenAI Stack — model-agnostic, OpenAI-compatible agentic API server for any model and infrastructure.
-
Lola
AI context package manager — portable skills, agents, and AGENTS.md modules for cross-assistant deployment.
Authors
Articles are written by the host author and invited collaborators. Each byline links the piece to the person who wrote it; guest posts appear with the contributing author's profile.
Vincent Caldeira
Red Hat APAC CTO · Columbia University Visiting Scholar · AAIF Ambassador
Host · AAIF Ambassador Cohort 2026
Vincent Caldeira, Red Hat APAC CTO and Industry Visiting Scholar at Columbia University, drives tech strategy and emerging technology engineering. A Top 10 APAC CTO (2023) with 20+ years in finance IT, he is an authority on open source, cloud-native technologies and AI. Vincent holds leadership roles across the Linux Foundation (FINOS, LF AI & Data, Agentic AI Foundation, CNCF) and GSF, championing AI safety, sustainability, and digital sovereignty.
Christopher Nuland
Chief AI Architect, Red Hat
Contributing Author
Christopher Nuland serves as the Chief AI Architect at Red Hat, specializing in AI adoption, Inference, Agentic, cloud-native migrations, AIOps, and metrics-driven transformations. He brings a robust background in leading machine learning and big data analytics initiatives across various industries and verticals. Currently, his research focuses on semantic routing in enterprise configurations. Christopher is actively involved in the PyTorch community and serves as a governing member of the vLLM Semantic Router community. A prominent keynote speaker, he has shared his expertise at global conferences such as KubeCon US/EU, IBM Think, and PyTorch Con.
Brand Mascot
Vince the Validator
Publication brand emblem
Vince the Validator is the visual emblem of thinkinginsidethebox.ai — a minimalist robot who guides agent reasoning safely inside the Box. The mascot represents the publication's brand identity and is distinct from the authors behind the articles.