Coverage
Every knowledge point in the model, and whether the question bank writes about it.
Numbers generated 2026-10-27T18:53:46Z from pipeline records; first source digest 97a3df0b. A stale dashboard gets quoted, so the page states when it was built.
How this figure is computed, and why it is lower than a count of mentions.
A knowledge point counts as covered when a question names its concept — the join above, through 225 concepts. Counting points a question merely mentions in passing gives a much higher number, and that is the figure this project has quoted elsewhere as the headline. Both are true; only this one answers “could I be asked about it”.
It does not measure whether the question is good. Every item in the bank is ai-reviewed and none has been read by a person, so a covered point means a question exists, not that it is a fair one. Working the review queue is what changes that.
Why 205 knowledge points have no question
All 205 of the 288 TP 15263 syllabus objectives have none - statements like "describe", "explain", "list" rather than propositions, and not one of them carries a citable sentence. Questions were written where a passage exists to serve as an answer key: the regulation-derived knowledge points are 602 of 602, every one covered. The gap is not difficulty and not neglect - it is what a defensible multiple-choice question can be built on.
Evidence pipeline: 122 objectives still unresolved
- 5draft-eligible - a confirmed passage, ready to draft against
- 53a relevant sentence exists but does not settle the question
- 122nothing left to read - a source must be acquired
- RPA airframes, power plants, propulsion and systems59
- meteorology27
- radiotelephony12
- navigation9
- theory of flight6
122 drafts exist and none is validated; 117 passed a blind re-check, which is a model check, not a review. Nothing on this page counts as coverage until a person approves an item.
By domain, least covered first
A domain links to a practice run filtered to it. The filter joins through the model — a quiz item names a concept, and the concept’s knowledge point names the domain — which is the same join that produces the Covered column, so the link and the figure cannot end up describing different sets of questions.
25 of 901 knowledge points carry no domain, so this table describes the 876 that do. Their concepts cite no regulation part to file them under, so they are left unfiled rather than given a plausible-looking bucket — which is why the number is stated instead of the rows being quietly dropped. It was 620 before the rest were filed by the instrument each point cites.
A domain is a part of a document, not a lesson: CAR 901 is Part 901 of the regulation, TP 15263 area 2 is area 2 of that knowledge guide. Read a row as 'of the N points this part of the document is responsible for, M have a practice question'.
If you are preparing for an exam: the covered column is the material you can drill, and a row below 100% is material you can only learn from the lessons. It is not a prediction of what will be asked.
| Domain | Covered | Of | Coverage | Practise |
|---|---|---|---|---|
| TP 15263 area 2 | 8 | 99 | 8.1% | Practise → |
| TP 15263 area 4 | 9 | 48 | 18.8% | Practise → |
| TP 15263 area 5 | 8 | 33 | 24.2% | Practise → |
| TP 15263 area 6 | 9 | 31 | 29% | Practise → |
| TP 15263 area 8 | 2 | 6 | 33.3% | Practise → |
| TP 15263 area 3 | 5 | 14 | 35.7% | Practise → |
| TP 15263 area 7 | 9 | 14 | 64.3% | Practise → |
| TP 15263 area 1 | 26 | 36 | 72.2% | Practise → |
| CAR 901 | 322 | 322 | 100% | Practise → |
| CAR 921 | 15 | 15 | 100% | Practise → |
| CAR 922 | 29 | 29 | 100% | Practise → |
| CAR 900 | 33 | 33 | 100% | Practise → |
| CAR 903 | 30 | 30 | 100% | Practise → |
| no domain recorded | 25 | 25 | 100% | Practise → |
| AIM RPA 3-2 | 66 | 66 | 100% | Practise → |
| AIM RPA 3-4 | 12 | 12 | 100% | Practise → |
| AIM RPA 2-0 | 4 | 4 | 100% | Practise → |
| AIM RPA 1-0 | 7 | 7 | 100% | Practise → |
| AIM RPA 3-6 | 7 | 7 | 100% | Practise → |
| AIM RPA 3-3 | 3 | 3 | 100% | Practise → |
| AIM RPA 3-1 | 3 | 3 | 100% | Practise → |
| AIM RPA 3-0 | 4 | 4 | 100% | Practise → |
| Transport Canada guidance pages | 10 | 10 | 100% | Practise → |
| CAR 301 | 7 | 7 | 100% | Practise → |
| CAR 602 | 26 | 26 | 100% | Practise → |
| CAR 103 | 5 | 5 | 100% | Practise → |
| CAR 302 | 9 | 9 | 100% | Practise → |
| CAR 601 | 3 | 3 | 100% | Practise → |
Questions about concepts the model does not file
3 concepts the bank writes about that reach no knowledge point. These are items testing something the model does not know it contains — a gap in the model rather than in the bank, and listed here so it is a known gap rather than an unexplained difference.
concept:registration.proof-of-registration— Registration proof of registrationconcept:tp15263.a2.5.batteries.discharge-curves— Discharge curvesconcept:basic-operations.division-iv-application— Basic operations division iv application
Edges by type
Every edge in the model, by type, with whether a person has confirmed it and whether it carries the evidence quote an admitted edge requires. 951 edges across 7 types.
| Edge type | Count | Human-confirmed | Evidence | Gates a learning path? |
|---|---|---|---|---|
regulated_by | 327 | not required | 327 / 327 | no — structural |
tested_by | 283 | not required | 283 / 283 | no — structural |
cross_reference | 150 | not required | 150 / 150 | no — structural |
related_to | 97 | not required | 97 / 97 | no — structural |
prerequisite_of | 92 | 0 of 92 | 92 / 92 | yes — sequencing |
exception_to | 1 | not required | 1 / 1 | no — structural |
applies_to | 1 | 0 of 1 | 1 / 1 | yes — sequencing |
0 of 93 sequencing edges are human-confirmed. prerequisite_of and applies_to are never machine-admitted, and neither may gate a learning path until a person confirms it. The 93 edges of those types are therefore carried by the model and used by nothing that routes a learner: practice ordering treats an unconfirmed edge as evidence of nothing.
Two metrics this page cannot compute, and why. A breakdown by tier and an auto-admit rate are both asked for by the design. The export stores no tier field — the three confidence tiers describe the design, they are not a column — and it does not record which gate step admitted each edge, so a tier count or an auto-admit percentage would be invented here rather than derived. What the model does record is reported instead: 951 of 951 edges carry an evidence quote, and every edge’s creator is named below.
Provenance across 9 creators. 31 edges name no creator, which is a defect in the extraction rather than a gap in the model. derived:tp15530 311 · agent:phase2-fanout 281 · derived:tp15263 207 · agent:graph-backbone 91 · agent:phase2-batches 23 · unattributed:phase1 21 · unattributed:phase0 10 · derived:delegation 5 · agent:graph-prereq-miner 2
Browse the edges
Pick a concept to see the typed relations the model holds about it, each with the passage that justifies it. Proposed sequencing edges are withheld — a reviewer is where those belong — and the number withheld is stated rather than left to be noticed.
The graph is typed edges over the concepts. Browsing it downloads the whole model file, so it loads only if you ask for it.
The model, as it stands
| concepts | 537 |
| knowledge points | 901 |
| edges | 951 |
| sequencing edges confirmed by a person | of 93 — the ranking in practice treats unconfirmed edges as evidence of nothing |
| knowledge points adopted by a person | 0 |