Stage 9advanced28 min
Building Agents
Put it together: architecture, instructions, tool design, evaluation, observability and safe autonomy.
Start with Architecture firstLessons
- Architecture firstMost agent projects fail at the design stage, by choosing an agent for a task that needed a workflow.7 min
- Writing agent instructionsInstructions that survive a long, messy run look different from a good chat prompt.7 min
- Designing tools and limiting blast radiusThe tools you provide define what the agent can do, and therefore what it can do wrong.7 min
- Evaluation and observabilityYou cannot improve what you cannot measure, and agent runs are unusually hard to see into.7 min