We Measured Where GPT-4o Breaks — And Built the Kill Switch
Validating the Deterministic Horizon on real, free infrastructure — and binding a model's measured reasoning limit into a cryptographic Depth Covenant. Includes a live demo.
Field Notes
Writing on accountable agents.
Practical notes on cryptographic identity, signed Agent Cards, evidence exports, and the EU AI Act readiness work behind agent-seal.xyz.
Agent Seal papers
The updated repository paper and the standards position paper, written for people who need to understand what the registry proves and where its legal boundary sits.
Pilot case studies for accountable agents
Two practical records: a support disclosure agent and an Annex III-style benefits triage agent. Each shows what the public evidence proves and what the provider still has to own.
Agent-Seal.xyz: A Production-Ready EU AI Act Evidence Layer for Autonomous Agents
Why autonomous agents need identity, keeper accountability, tamper-evident traces, public status, and court export before high-risk deployments reach audit pressure.