Bound the assignment, not the answer
Every assignment supplies a role, cutoff, decision, source neighborhood, and handoff. Learners still own definitions, methods, uncertainty, and recommendations.
PEDAGOGICAL OPERATING STANDARD
Practice judgment before the stakes rise.
The Analyst is the simulation; Meridian Living Systems is the fictional employer that makes the work cohere. Each assignment changes the role and decision while the organization, source systems, definitions, and operating history remain consistent. The world is controlled enough to teach from, but the analytical choices remain real. It can support college and university courses, workforce programs, professional training, or structured self-study.
Make Python earn its place
SQL establishes governed evidence. Python handles work that benefits from programmatic analysis: visualization, inference, temporal validation, simulation, modeling, and error analysis.
Preserve useful mess
The estate contains declared source behavior—multiple clocks, sparse observations, fanout risk, delayed labels, technical replays, selection effects, and shifting populations. These are mechanisms, not random corruption.
Permit responsible refusal
A defensible result may be hold, monitor, shadow, collect better labels, or decline deployment. The simulation never requires false certainty to feel complete.
Separate mechanics from judgment
Machines may verify execution, file integrity, grain, referential constraints, and exact authored invariants. Instructors evaluate analytical meaning, assumptions, causal language, risk posture, and decisions.
Leave work another analyst can inspect
The submission contains code, queries, notes, evidence, outputs, run records, data revision, runtime versions, and content hashes. Reproducibility is part of the work product.
MACHINE-VERIFIABLE
Runs, schemas, file hashes, declared grain, exact row constraints, referential integrity, temporal cutoff violations, and scenario-authored invariants.
HUMAN-REVIEWED
Question framing, statistical reasoning, model choice, uncertainty, business consequences, causal restraint, communication, and whether the recommendation deserves action.
SCAFFOLD MODES
Templates, small fixtures, query contracts, fold utilities, and explicit reconciliation prompts.
Data neighborhood, cutoff, deliverable contracts, and hints for mechanically detectable failure modes.
Executive request, source estate, workbench, and handoff requirements. The analytical route remains learner-owned.
TEACHING AND ASSESSMENT
OPEN INSTRUCTOR DESK →The low-profile instructor desk holds sequencing guidance, the human-review boundary, a local submission viewer, and spoiler-separated assignment notes. None of those materials are required for self-directed use.