DrCatHicks/learning-opportunities

What it solves

It addresses the "fluency illusion" and learning decay associated with agentic AI coding. When developers rely on AI to generate code, they often skip the active mental processing required to truly understand the architecture and logic, leading to a decrease in long-term skill development and retention.

How it works

This project provides plugins for Claude Code and Codex that introduce a "dynamic textbook" approach to coding. After a developer completes significant architectural work (like schema changes or refactors), the AI prompts the user to engage in a 10-15 minute learning exercise.

Unlike standard AI interactions, the system intentionally pauses and waits for user input to force active retrieval and generation. It employs several evidence-based techniques:

• Prediction and Reflection: Predicting an outcome, observing it, and reflecting on the difference.
• Generation: Sketching an approach before seeing the AI's implementation.
• Retrieval Practice: "Teaching back" a component or performing a memory check-in at the start of a session.
• Tracing: Walking through execution steps manually.

Who it’s for

Developers using agentic AI coding tools who want to maintain or build their technical expertise while increasing their production velocity.

Highlights

• Science-Backed: Based on cognitive psychology principles like the spacing effect and the generation effect.
• Integrated Workflow: Plugs directly into Claude Code and Codex as a marketplace plugin.
• Repo Orientation: Includes an orient skill to help developers strategically navigate and learn new codebases.
• Automatic Prompting: Optional hooks for Linux/macOS to trigger learning prompts after git commits.