Loading local project records…

Course 12 project

A world model that earns its use

Freeze a training recipe and planner, compare prediction with externally executed control, and decide whether the model earns its additional work. Review the evaluation lesson or open the implementation bridge. All routes remain freely accessible.

1. Freeze and collect evidence

Current freeze: none. Known reserved exposure: 0. Freezing again never clears the first reserved recipe or assessment attempts.

The public reserved starts differ from development and training starts, but their outcomes are inspectable in source. Repeated exposure is practice. Both nominal and changed mass/drag cases remain in each capture. Success requires all 20 intervals and final |x|<0.2 m, |v|<0.3 m/s; imported scores are ignored and recipes replay.

2. Explain the decision

Guided: explain the comparisons. Decision: justify accepting, limiting or rejecting model-based planning. Implementation: attach your source/tests and independently replay changed starts using the studio. The code field is inert; a supplied reference run does not establish implementation proficiency.

Outstanding: Capture a frozen development comparison. Capture a reserved comparison after freezing. Capture development and reserved evaluations with the same frozen coverage and planner horizon. Write prediction. Write uncertainty. Write control. Write budget. Write decision. Write limits.

Report not submitted. Submission checks presence and recipe evidence, not reasoning or source quality. Human review is pending.

3. Separate transfer checks

These scalar cases differ from the lab. Opening and all first answers persist. Three submissions maximum; retries are practice. Public client-side questions do not guarantee secrecy.