Research
Long-form investigations into agent architectures, risks and operational realities.
Our mission
We study how AI models become agent systems and how those systems are operated responsibly — across harnesses, runtimes, control planes, governance, permissions, memory, execution, observability and recovery.
We publish what we learn, including what doesn’t work.
Independence
We are not owned by or affiliated with any vendor.
Evidence first
Claims must be backed by reproducible evidence.
Transparency
We document methods, data, versions and limitations.
Human review
All published work is reviewed by experienced practitioners.
Open by default
We share data, code and methodology whenever possible.
Service to the community
We aim to raise the standard for the entire ecosystem.
What we do
Research, evaluate and explain agent systems.
Experiments
Reproducible benchmarks and field tests across models, tools and runtimes.
Analysis
Clear, critical analysis that surfaces trade-offs, failure modes and mitigations.
Education
Guides, explainers and briefings to help teams operate responsibly.
Who we work with
We collaborate with engineers, security teams, product leaders and executives.
Engineering teams
Building and operating agent systems.
Security & GRC
Managing risk, compliance and auditability.
Product leaders
Evaluating vendors and setting strategy.
Researchers
Advancing the science and practice together.
- Agent architecture
- Governance
- Evaluation
Founder
Adam Maguire Wilson
Founder and Principal Advisor
Adam Maguire Wilson is an independent advisor and researcher focused on AI agent systems. He advises engineering, security and product teams on agent architecture, governance and operations, and publishes his ongoing research here and in the Agent Field Notes briefing.
Our editorial commitments
Accuracy
We strive for factual accuracy and correct errors transparently.
Fairness
We evaluate systems fairly and avoid cherry-picking.
Context
We provide the context needed to interpret results and limitations.
Accountability
We stand behind our work and welcome constructive critique.
We’re always learning.
Follow our work, subscribe to Agent Field Notes, or get in touch.