- Agent architecture
- Governance
- Evaluation
Author
Adam Maguire Wilson
Founder and Principal Author · Harness the Agents
Adam Maguire Wilson founded Harness the Agents and is its principal author. He writes most of the publication’s technical work: hands-on examinations of agent harnesses, runtimes and control planes, pinned to specific versions and published with methods and limitations on the record.
His focus is how AI models become dependable agent systems — architecture, permissions, memory, observability and recovery — and how the teams running them govern those systems in production.
He also publishes the Agent Field Notes briefing. His personal site — broader writing and research beyond this publication’s technical register — is adam.mw.
Latest writing
Every article is installed, read and run before it is written up — see the methodology for how examinations work.
Amodei's Embedded Evaluators: Access Is the Whole Game
Amodei pledges embedded evaluators with employee-like access and publication rights. What the pledge promises, what it leaves open, and what is on paper today.
An Agent That Works for Weeks Needs More Than Memory
Salesforce's long-horizon runtime lets an Agentforce agent pursue a goal for weeks. Memory, durable execution and dynamic steering are documented. The authority questions are not.
Anthropic's Threat Report: The Shift That Matters Is the Workflow Around the Model
Anthropic's September 2026 threat report shows misuse industrialising around the model: orchestration, retained context, parallel work. What operators must change.
OpenAI Now Sells the Agent Harness. You Still Own the Consequences.
OpenAI's Agents API puts the Codex harness behind one API call. The docs are candid about what the managed harness does, and what remains your problem.
Your Agent Can Fail Without Anyone Attacking It: Cisco's New Taxonomy Names How
Cisco's framework v2 names how agents fail with no attacker: excessive agency, goal drift and reward hacking, and why each failure needs different controls.
OpenAI's Navier-Stokes Swarm: Verification Is a Separate Job
OpenAI says 10,000 agents produced a Navier-Stokes blowup proof in 88 hours. What the Lean certificate checks, what it cannot settle, and the lessons for agent builders.