Independent project · September 2026

Recall

Focused practice for adult learners, shaped around clear guidance, trustworthy feedback and later review.

The project

Recall helps adult learners turn difficult study material into focused practice and later review.

My contribution

I led the product direction, identified weaknesses in the learning experience, and guided AI-assisted implementation and testing using Codex.

The constraint

AI-generated practice needed to give trustworthy feedback. A card saying “Needs matter” could correctly belong to both conduction and convection, but the activity accepted only one answer. A learner could reason correctly and still be marked wrong.

The decision

I decided to reject ambiguous cards during generation and block affected older saved practices before scoring. Learners could create a replacement without losing the original. The trade-off was stopping an affected activity instead of letting it continue. Fair feedback mattered more than keeping every activity available.

The result

The older ambiguous practice was blocked before scoring in both full practice and scheduled review. These checks established release readiness; real learner testing is still needed to assess usefulness and return usage.

Technology

Next.js
React
TypeScript
Tailwind CSS
shadcn/ui
Vitest

Engineering notes

The detail behind the build.

Evidence & verification

I combined automated regression tests with browser testing on preview and production. I tested pasted text, PDF extraction, passage selection, practice, Memory Helper, retries, and saved-data persistence. After deployment, I confirmed that the older ambiguous practice was blocked in both full practice and scheduled review.

Reflection

Technical correctness alone does not make a learning product compelling. Learners also need clear guidance, trustworthy feedback, and an honest distinction between answering after help and remembering independently.