Technical guides to the systems behind AI-native businesses. Architecture, controls and methods, from first principles to production decisions.
001 / Security engineering
Secure the entire agent loop: identity, prompt injection, tool execution, human approval, data boundaries, audit and adversarial testing. Ten layers, worked scenarios and a release checklist.
Read the blueprint002 / Quality engineering
Know whether an agent is ready to ship: success criteria, real-work test sets, graders, trajectory checks, red teaming, regression gates and production monitoring.
Read the blueprint003 / Operating design
Decide what agents do alone and what stays with people: autonomy levels, meaningful approvals, escalation, reviewer screens, override controls and accountability.
Read the blueprint004 / Operations engineering
See what agents did, why and on whose authority: run capture, tracing, append-only audit trails, privacy in logs, metrics, behavioural alerts, replay and reporting.
Read the blueprint005 / Harness engineering
Engineer the system around the model: context, tool contracts, controlled execution, durable state, isolation, approvals and evidence-based completion. Ten layers, failure scenarios and a production checklist.
Read the blueprint