Writing & guides

Notes from
doing the work.

Practical explanations, engineering decisions, and experiments with enough detail to follow the reasoning.

Read the essays ↓Browse the writing archive →

Essays & field notes

From systems in practice.

Agents, retrieval, evaluation, and the decisions that make production AI reliable.

Apr 30, 20264 min

MCP is not a tooling problem. It is a governance problem.

After shipping 458 governed tools across an MCP platform, the lesson isn't about volume. It's about the four production failure modes that nobody writes blog posts about — and why most teams ship MCP the same way they shipped REST APIs in 2009.

MCPagentsgovernanceproduction AI

Keep exploring

Guides & research.

Free field guide

Learn by building the loop ↗

Eight practical missions covering agents, prompting, retrieval, evaluation, and tool use.

Research record

What the Wisp experiments showed ↗

Methods, comparisons, null results, and the limits of a small code model.

The writing archive

49 dated entries across this site, DEV Community, Medium, and LinkedIn.

Notes fromdoing the work.