Free Guides
Library
Practical guides for building dependable AI agents — clear enough for operators to follow, structured enough for agents to execute, focused on the systems work that holds up in production.
Start with any guide below. Each one is designed to stand alone, but together they form a coherent operating model.
Pre-Mortems for Agent Tasks
A focused technique for catching foreseeable failures before a multi-step agent task runs. One method, applied well, prevents a class of avoidable mistakes.
Memory Architecture Patterns
Why flat memory breaks at scale, and how a three-layer architecture fixes it. A concrete pattern with file layout, decision rules, and clear tradeoffs.
Multi-Model Review: A Practical Guide
When to send output to a second model, how to structure the review prompt, a severity-tiered rubric, and an honest cost/benefit accounting.
Scheduling Architecture for Autonomous Agents
Most agent scheduling is stateless — run this every N hours, no learning, no cost optimization. The fix is treating scheduling as an architecture problem, not a config problem.
Self-Modification Under Governance
The debate is 'should agents self-modify?' That's the wrong question. The right question: which modifications are safe to automate, and what governance makes the rest tractable?
Want the complete system?
Blueprint turns these patterns into a working operating stack.
Each guide covers one capability. Blueprint wires them together — memory, orchestration, safety boundaries, and execution standards — into one self-executing setup your agent ships in under an hour.