The question: Away from the original example and model answer, can you apply the method to a new problem?
Before you begin: Complete an action card and try it in a real low-risk task.
- 01Unaided recall
- 02Find the gap
- 03Transfer the idea
- 04Retest later
Concepts and learning objectives
Understanding an explanation is not the same as independent ability. Practise closed-book explanation, repair and transfer to a new setting. Identify inputs, boundaries, failures and verification again. AI may ask questions and provide feedback after you answer, rather than reveal the solution first.
Day-seven and day-21 retests are suggested course rhythms, not universal optimal memory intervals. Review earlier learning now and schedule later checks for current material. A knowledge base also needs subtraction: remove duplicate, unsourced and obsolete cards without historical value.
Reading and selected sections
| Priority | Source and selected sections | Budget |
|---|---|---|
| Required review | R45 · Anthropic — How AI assistance impacts the formation of coding skills — Separate task completion from independent understanding; avoid broad claims from one study. | 30 min |
| As needed | R27 · MDN — JavaScript Guide — Return to functions, async code or error handling where weaknesses appear. | 45 min |
| Foundation | R16 · Harvard CS50x 2026 — Only the algorithm or data-structure concepts you cannot yet explain. | 45 min |
Studio lab
- Answer three earlier questions without a generator: explain a concept, find a counterexample and describe an application. Preserve first answers separately from later research.
- Transfer one principle: calendar deduplication to document import, or API error states to file handling. Explain differences as well as similarities.
- Make a small change independently, then let the model review and question it. If the model replaces your entire answer, shrink the exercise.
- Review cards for actual use, evidence and actionable next steps. Merge duplicates, mark version scope and delete unused generated summaries.
- Schedule a few important retests. Count maintenance of the learning system itself in your weekly review.
Deliverables and acceptance
| Deliverable | Evidence of completion |
|---|---|
| Independent retest | First answers and corrections are distinct. |
| Transfer application | The principle works in a new scenario and changed boundaries are explained. |
| Knowledge-base cleanup | Clear reasons for deletion or consolidation; collection size is not the metric. |
Independent exercise: Repair a small unseen problem without a generator. Official documentation is allowed; explain why the fix works.
Privacy and safety: Use no other person's private information in exercises. Follow the originating course's academic-integrity rules, including R15. Keep your own AI-assisted projects separate from restricted coursework.
Everyday use: Notice whether you repeat fewer searches, avoid rework or recognise a risk sooner. Needing more foundation practice is a legitimate result.
Pass or revisit: Demonstrate one independent transfer before increasing deployment complexity. Delay week 15 if the existing tool is still unexplained.
A reusable collaboration prompt
Be a questioner, not an answer generator. Give new scenarios for three concepts and wait for my answers before feedback. Help consolidate useless notes and schedule a few retests without building a large learning-management system.
Sources and verification scope
Reference IDs and scope follow the supplied manuscript, baseline 2026-09-27. External authenticated labs were not all completed; recheck versions and access before implementation.
- R15CS50P — Academic Honesty ↗
Formal assignments follow the current policy; personal projects are separate.
- R16Harvard CS50x 2026 ↗
Select algorithms, memory and data structures according to gaps.
- R27MDN — JavaScript Guide ↗
Syntax, control flow, collections, functions, async and modules.
- R45Anthropic — How AI assistance impacts the formation of coding skills ↗
Controlled study with bounded sample and tasks; not a claim about all AI-assisted learning.