Lab 1: Manual tool-use loop
Build the application-controlled loop behind a small note assistant, then make every stop state visible.
Loading…
Hands-on practice
Open-ended assignments for practicing architectural judgment.
Choose the local architecture path or build with your own tools. Nothing is uploaded.
Marking a lab complete records your progress. It isn't a grade.
decision.mdState the requirement, chosen pattern, rejected alternatives, and the reason for the tradeoff.
implementation/Keep the smallest working example, configuration, diagram, or synthetic-trace analysis.
evidence.mdRecord test cases, observed results, failure behavior, limitations, and evidence locations.
review.mdExplain what must be deterministic, what may remain model-directed, and where a human is required.
Create a folder for the lab, use fake data, and start with these files. The architecture path doesn't require an API account.
Labs
Build the application-controlled loop behind a small note assistant, then make every stop state visible.
Coordinate three bounded specialists without losing provenance or hiding a coverage gap.
Enforce safety policy outside the prompt and produce a handoff another person can act on.
Compare continuation and branching while keeping conversation state distinct from workspace state.
Design narrow MCP capabilities for a synthetic note catalog and test every trust boundary.
Place guidance, reusable procedures, scoped rules, and enforcement in the mechanism that fits each one.
Make a review job distinguish a clean diff from malformed output, execution failure, and invalid evidence.
Compare extraction strategies against labeled synthetic data without hiding critical-field failures.
Decide what stays active, what becomes durable, and what must be revalidated after recovery.
Evaluate autonomy, escalation, failure propagation, and claim-level provenance across a synthetic scenario suite.
Synthesize one fictional system into a coherent, reviewable architecture packet.