Release record / engine v1
How every worksheet
earns release.
A worksheet is not released because it looks plausible. Its settings must match a published preset, its seeded run must complete, its stored answers must agree with a fresh recomputation, and the final record must pass the checks the code can actually perform.
Inspection station 01
The job begins as a bounded specification.
The builder accepts a versioned set of choices. Grade, skill, variant, difficulty, length, paper, layout, workspace, and answer-key mode travel together. The selected preset supplies the mathematical constraints; a mismatched preset is rejected before generation.
A valid specification defines what may run. It does not, by itself, prove instructional quality.
schema 1 / engine 1- Grade
- K, 1, 2, 3, 4, 5
- Difficulty
- Easy, standard, challenge
- Problem count
- 10, 20, 30, 40, 50
- Paper
- US Letter or A4
- Orientation
- Portrait or landscape
- Print size
- Standard or large
- Workspace
- None or answer lines
- Answer key
- None or appended
Inspection station 02
The same release record can be replayed.
Engine version, preset id, and worksheet seed are joined before the deterministic random stream begins. The same specification, preset, seed, and engine version produce the same ordered problems. Change one part and the output may change.
The seed is a replay handle, not a security token. Random selection chooses operands and formats; exact answer arithmetic is handled separately.
1:3-multiplication-facts-within-100:release-record-example- 01Version + presetFix the generator contract.
- 02Worksheet seedFix the replayable run.
- 03FNV-1a initialization
0xdf10c494 - 04Candidate streamReject, retry, and dedupe deterministically.
- 05Ordered set10 problems in this record.
Inspection station 03
A stored answer must meet a fresh recomputation.
Each generated problem keeps a structured computation record and an exact stored answer. During validation, the computation record is run through the deterministic recomputation path again. The resulting exact value is compared with the stored value, and the printable answer text is checked against its formatter.
Important: this is a fresh run through the same computation model, not a separately implemented or external solver.
integerInspection station 04
Release is a matrix, not a badge.
Different checks happen at different points. Generation enforces preset-specific candidate constraints. Final validation checks the completed problem set. The PDF renderer refuses a worksheet that does not pass that final validation.
The table names the implemented path and its evidence without turning test coverage into a promise that every pedagogical decision is correct.
Swipe the matrix to inspect evidence and failure behavior
| Gate | What the code does | Evidence in this project | Failure behavior |
|---|---|---|---|
| 01Specification | Accepts only versioned settings and requires the grade, skill, and variant to match the selected preset. | isWorksheetSpec and preset-mismatch tests. | Generation stops with an explicit invalid-spec or preset-mismatch error. |
| 02Constraints | Applies normalized preset and difficulty constraints while candidates are created. A candidate that violates a supported range or format is rejected and the seeded stream retries. | Semantic catalog tests plus standard and 50-question generation across 81 leaf presets. | After a bounded retry limit, generation stops rather than returning an undersized set. |
| 03Dedupe | Skips repeated candidate fingerprints during generation, then scans final dedupeKey values again during validation. | Unique-set assertions and a final duplicate scan in validateWorksheet. | A repeated candidate is skipped; a repeated problem in the final set invalidates the worksheet. |
| 04Answer format | Recomputes the structured computation, compares the exact answer, and confirms that printable text matches the exact stored value and display mode. | Tampered-answer detection and exact integer, fraction, decimal, money, and time tests. | An incorrect value, invalid computation, or mismatched answer string invalidates the worksheet. |
| 05Print release | Runs final worksheet validation before creating a selectable-text PDF, then applies the requested paper, orientation, print-size, and answer-key mode. | PDF parse/page-count tests, A4 landscape dimensions, answer-key modes, and specialized visual-render tests. | The renderer refuses an invalid worksheet. Rendering errors are reported instead of silently shipping a file. |
Inspection station 05
Sources and corrections remain attached to the record.
Standards labels are editorial alignment notes. Each reference carries its source, version, alignment role, and verification date. We link to the issuing body and keep that source record separate from our own summary.
- Source version
- Official web edition, accessed 2026-07-15
- Record verified
- Worksheet use
- Concise primary or supporting alignment note—not a claim that one printable fully assesses the standard.
Correction loop
A reproducible report goes back through the same gates.
- 01Record
Send the page address, worksheet seed, and question number.
- 02Replay
Recreate the same specification and ordered set.
- 03Isolate
Separate a computation, constraint, wording, source, or rendering issue.
- 04Correct
Change the relevant code or content and rerun the release checks.
- 05Record again
Regenerate affected samples and update the check date for material changes.
Release the next practice set
Choose the bounds.
Inspect the proof.
The builder exposes the settings and worksheet preview before you create the printable. Start with a grade and skill, then keep this record nearby when you want the details behind a check.
Methodology record last updated .