Posts I've written and other content I’ve contributed to.
MCP is marketed as the USB-C of AI integrations — but plugging raw upstream schemas directly into an agent's context window is a textbook DDD anti-pattern. Here's why, and what to do instead.
A weekly live rundown of the most important AI news, filtered through the lens of people who build with this stuff every day. Hosted by Thoughtworks' Ben O'Mahony and Fabian Nonnenmacher, with guest Rishi — covering Gemini 3.5, recent model releases, and what actually matters for teams shipping AI products.
Ben O'Mahony and Fabian Nonnenmacher join host Ken Mugrage on the Thoughtworks Technology Podcast to discuss their O'Reilly book "Building AI Agent Platforms" — covering what agents really are, how to design a platform for them, evaluations, observability, guardrails, and the role of reference implementations.
A blueprint for building AI agent platforms that scale. Covers the full agent lifecycle — from deployment to evaluations — plus MCP for agent interoperability and the enterprise challenges of security, auditing, and governance.
Fabian Nonnenmacher demos a lightweight GenAI agent that connects to Jira, epics, and documentation to overcome the blank-page problem when drafting user stories. Includes a practical walkthrough with Streamlit, PydanticAI, and Cursor.