Create approved flashcards from a learner’s own knowledge gaps

Proposed inputs

1. Name

   Course evidence

   Owner

   Learner or educator

   Minimum scope

   A short owned excerpt and authoritative answer notes.

   Quality check

   Keep page locators and distinguish a corrected answer from an unresolved disagreement.

2. Name

   Gap decisions

   Owner

   Learner

   Minimum scope

   Responses for this excerpt and a selected destination.

   Quality check

   Self-confidence is provisional; do not infer a grade or mastery from it.

Access and setup — proposed or unknown

Proposed minimum

1. Start with an offline proposal and a disposable profile; inventory any reused note-type names before later setup.

Compatibility unknowns

1. Unknown: installed host/add-on compatibility and effects on the owner’s customized note types.

External actions requiring approval

1. Approve content processing, exact notes and destination; approve template/style changes separately from adding notes. Do not sync or alter scheduling as a side effect.

Why use this

1. More cards create a review obligation. The useful question is which missing distinction is worth practicing, rather than how many slides can become questions.

2. A compact approval sheet lets the learner reject misleading cues before they become recurring homework.

Why an agent / simpler alternative

Useful judgment

Compare an attempted answer with an authorized explanation and propose a narrowly scoped recall target.

Simpler alternative

Manually add a few known gaps when the source and intended question are obvious.

Unsuitable when

The deck contains sensitive learner records, unsupported answers or high-stakes professional advice.

Expected deliverable and review checks

Proposed artifact

An itemized proposal showing target, cue, answer, source, decision and expected note/card counts.

Critical checks

1. Every answer is grounded in the selected excerpt; familiar targets can be omitted.

2. Count notes and generated cards separately, since one note may create several cards.

3. A later authorized import must reconcile individual returned IDs with approved items and preserve unrelated templates.

Missing output evidence

No completed output or independently observed run was inspected in this research.

First check — proposed, unexecuted

Status

proposed_unexecuted

Synthetic fixture

Invent a two-paragraph lesson about a fictional garden-label system. Provide four responses: two sound, one ambiguous and one mistaken; include an existing card for the mistaken topic.

Acceptance checks

1. The ambiguous response becomes a clarification, not an assumed deficit.

2. The existing card is flagged for a learner decision; proposed additions exclude already-known targets.

3. No collection write, model refresh or sync is part of this check.

Stop conditions

1. Stop if answer authority, destination or preservation boundaries are unclear.

Failure signals and proposed responses

1. Signal

   A card rewards a section number rather than the idea.

   Why it matters

   The cue can test book navigation instead of understanding.

   Proposed response

   Rewrite against the learning target or omit it.

2. Signal

   A failed batch may already have added some notes.

   Why it matters

   Blind retries can duplicate work.

   Proposed response

   Inspect existing IDs and reconcile the approved set before proposing any retry.

Effort and cost — unknown

Setup work

1. Choose a source-of-truth answer and prepare an isolated collection snapshot.

Ongoing review

1. The learner reviews wording, relevance and any disputed fact.

Unknowns

1. Unknown: review burden, model cost and retention benefit.

Limits

1. This is a proposed selection and review method, not a measure of learning gain. Source mutation caveats remain in the linked evidence.