17. Running a Fault-Injection Exercise for an Agent
A successful happy path proves that the system can run. A controlled failure shows whether it knows where to stop and how to recover.
Read noteBuilding systems,
one trace at a time.
A personal engineering notebook about AI systems, software, and the experiments that make them understandable.
Latest notes
Ideas, implementations, and details worth keeping.
16. Building an Agent Regression Suite Before Release
A regression suite is not a rerun of a few demos. It is an executable record of failures the system must not repeat.
Read note09. Agent Permission Is More Than Allow or Deny
Authorization decides whether one principal may perform one action on one resource under specific conditions
Read note07. Are Parallel Tool Calls Actually Faster?
Concurrency shortens only independent waiting time and adds result joining, cancellation, and resource pressure
Read note12. Context, Summaries, and Long-Term Memory Are Different
Longer retention does not create stronger knowledge; unverified memory carries old errors into new tasks
Read note15. Evaluating an Agent Beyond Its Final Answer
A plausible answer does not prove that evidence is real, execution was authorized, or interruption recovery is correct
Read note01. How System Prompts Are Assembled
The model does not receive one prompt. It receives context blocks assembled from different sources for different purposes.
Read note06. How Tool Results Should Return to the Model
The model does not observe local execution; the runtime must carry evidence into the next turn
Read note04. How Tool Schemas Change Agent Behavior
The model does not operate the system directly; it selects from the actions exposed by the runtime
Read note10. Preventing Duplicate Side Effects
Retrying a model call spends computation; retrying a send, payment, or create operation can change the external world twice
Read note