1 · Curriculum
Quality gates before quantity
Content must satisfy structural, coverage, rationale and mock-form checks before the production build can pass.
Transparent by design
How TaalKlaar decides what you should learn next.
A large question bank is not the method. TaalKlaar combines a versioned curriculum with your own retrieval, checkpoint and production evidence, then records whether each recommendation helped.
2 · Personal evidence
Your lesson has a reason
Due review, retained words, weak components, chosen focus and checkpoint timing determine the next session.
3 · Outcomes
Recommendations are testable
TaalKlaar records aggregate correctness, completion and time for each lesson decision—without making another copy of answer text.
Current production release
Auditable content coverage
A2
- Reading
- 300
- Listening
- 300
- Writing
- 76
- Speaking
- 108
- Grammar
- 18
- Mock forms
- 6
B1
- Reading
- 240
- Listening
- 240
- Writing
- 87
- Speaking
- 113
- Grammar
- 24
- Mock forms
- 6
- Unique item IDs and duplicate prompt/context checks
- Non-empty answer options, valid answer keys and learner-facing rationales
- Minimum topic coverage and six non-overlapping mock forms
- Required writing, speaking and grammar ranges enforced during the build
Important boundary
Evidence, not an official prediction.
TaalKlaar readiness is an internal learning guide. It does not use unpublished DUO data and cannot guarantee an official result. Versioned decisions make changes explainable rather than pretending the algorithm is certain.
Engine daily-plan-2026.07.1 · Rubric cefr-task-2026.07.1