SCIENTIFIC INTEGRITY · AUDIT 01
Biology that can show its working.
The complete curriculum now passes a repeatable integrity audit covering answer logic, explanations, evidence links, visual models, scientific guardrails and presentation quality.
✓ ZERO BLOCKING FAILURES IN THE CURRENT RELEASEEvery lesson interaction is inside the audit boundary.
Every choice and ordering task has mechanism-based feedback.
Every exercise resolves to a constrained biological model.
Every activity link was resolved to an existing research file.
CORRECTIONS FROM THIS PASS
The audit changed the learning material.
It is useful only when it catches and prevents real defects. These repairs were made in the source generator, so rebuilding the curriculum cannot silently restore the old problems.
- 24Explanations rewritten
Generic answer-key statements were replaced with question-specific biological or chemical reasoning.
- 3,360Provenance labels repaired
Damaged separator characters were corrected across every advanced activity.
- 158Long excerpts repaired
Corrupted truncation marks were replaced with readable punctuation.
RELEASE GATES
What must pass every time
The audit runs again whenever the application is validated, so later curriculum changes are checked against the same standard.
Curriculum completeness
96 units, 760 lessons, 4,200 activities and all deep-learning resources must remain reachable.
Answer integrity
Correct answers occur exactly once, sequences match their draggable steps, and assessed explanations contain biological reasoning.
Scientific guardrails
The audit rejects known category errors involving energy, evolution, antibiotics, circulation, excretion, enzymes, photosynthesis, gene expression and homeostasis.
Visual accuracy
All activities use one of 35 constrained visual families with an accuracy note and at least four explorable structures.
Evidence provenance
Advanced activities retain exact links to the source research layer, and every referenced local file must exist.
Presentation safety
Placeholder text, corrupted characters, duplicate choices and generic answer-key feedback fail the release.