James Mayo

AI systems architecture, agent design, conceptual modeling, systems thinking, adversarial testing, failure analysis, knowledge representation, tech writing, research synthesis, Python, JavaScript

Currently researching how long-lived AI agents can preserve continuity, recover safely, and expose failures hidden behind plausible outputs.

Currently hacking on systems that make AI agents more reliable—especially when a plausible final answer hides failures in how the agent got there.

I’m especially interested in connecting with people building long-lived AI agents—particularly around provenance, replay and lineage, authority boundaries, recovery, and adversarial evaluation.