A planner with uncertain observations.

Compare search methods and hidden-state models, then justify an observation or action decision. The workbench saves evidence and a report; four separate objective cases check transfer. Human artifact review remains pending.

Report: draft. Saved on this device in version 4 progress backups. Storage: loading. Export before clearing browser data or moving devices. Editing evidence or text reopens the draft and does not change lesson achievements.

1. Collect matched evidence

Save from the two lessons or import their exported JSON. Keep two search methods on the same map and budget, and two belief models on the same initial distribution and observation prefix. Six captures maximum. Full bounded search and filter traces are recomputed from settings; imported result fields are not trusted.

2. Explain the evidence

Describe the route and hidden-state questions. Identify public teaching data, capture IDs, budgets and shared observation prefixes.

Compare priorities, returned routes, exact verification, costs, reopenings and work. Distinguish a cutoff from exhaustive failure.

Work through one prediction and correction. Compare matched observation prefixes; explain uncertainty, missing readings and impossible evidence.

Inspect an overestimating search heuristic or excluded motion. Explain how a legal result or exact inference can still fail the intended task.

State priors, outcomes, action costs and observation cost. Compare both observation branches with the best direct action. Explain when you would defer or acquire more evidence.

State public exposure, untested cases, implementation limits and which claims need new data or human review.

Code is inert text and never runs here. Implementation proficiency requires runnable code, tests and independent review; the report presence check cannot establish it.

3. Submit the report

Submission checks required text and matched configurations. It does not evaluate reasoning, authorship, code quality or task success. Objective assessment below and human review remain separate; no new mastery or published-path requirement is awarded.

• Capture two search methods on the same map and expansion budget.
• Capture two belief models on the same initial distribution and observation prefix.
• Write the question report field.
• Write the search report field.
• Write the belief report field.
• Write the adversarial report field.
• Write the decision report field.
• Write the limits report field.