Reports

Long-form writing on AI systems work: what breaks after the demo, what to measure, and how to leave the system operable.

6 reports in Enterprise AI

Latest in this domain

Enterprise AISep 6, 20267 min read

AI agent kill switch governance: containment after rogue swarm reports

Autonomy ships before a kill switch, egress ACL, and reconstructable audit trail exist, so containment is improv when a swarm misbehaves.

Expose a privileged kill switch API that stops runs and tool sessions by durable run ID, not by hoping the agent cooperates.
Bound every tool-using agent with sandbox egress ACLs tied to a versioned allowlist, not open outbound networking.
Append audit events to an immutable hash chain so incident reconstruction does not depend on mutable application logs.

Enterprise AISep 6, 20268 min read
Data residency and model routing for enterprise AI platforms
One global model endpoint violates residency and vendor constraints
Read report
Enterprise AISep 6, 20267 min read
Policy-as-code for tool-calling agents in regulated workflows
Prompt-only policies drift and cannot be audited when agents gain write tools
Read report
Enterprise AISep 6, 20267 min read
Contract testing MCP tools: schemas, side-effect classes, and version skew
Tool schema drift breaks agents silently across environments
Read report
Enterprise AISep 5, 20267 min read
MCP enterprise gateways: auth, tool blast radius, and what breaks without them
One MCP server with a shared god key turns every confused agent into an unbounded actor.
Read report
Enterprise AISep 1, 20267 min read
AI gateway triple-gate: LLM, MCP, and A2A traffic under one policy plane
A gateway that only sees completions is blind to the actions that actually change systems.
Read report