The modern developer is currently living through a paradox of productivity. With the rise of agentic coding, a software engineer can now scaffold an entire feature, refactor a legacy module, or migrate a database schema in a fraction of the time it took three years ago. However, this velocity comes with a hidden cost. As AI agents take over the heavy lifting of implementation, the cognitive load shifts from creation to mere supervision. Many developers are finding that while their commit history is growing, their deep understanding of the codebase is shrinking. This gap between output and comprehension is what the industry is beginning to call the learning deficit.

The Mechanics of Active Learning

To combat this erosion of skill, GitHub has released a specialized learning reinforcement skills repository designed specifically for Claude Code and Codex. Rather than allowing the AI to silently handle every architectural shift, this tool intercepts the workflow to prompt the developer into 10 to 15 minute learning sessions. These sessions are not random; they are triggered by high-impact events such as the creation of new files, modifications to database schemas, or significant refactoring efforts. The system applies established principles of learning science, integrating techniques like prediction, generation, retrieval practice, and spaced repetition to ensure that the developer actually internalizes the changes the AI is making.

For those using Codex, the installation process requires cloning the repository and linking the package locally:

git clone https://github.com/p-m-p/learning-opportunities.git
cd learning-opportunities
npm install
npm link

Users of Claude Code can implement the tool via a global npm installation followed by a restart of the environment:

npm install -g learning-opportunities

For developers on Linux and macOS, there is an additional utility called `learning-opportunities-auto`. This tool automates the process by monitoring the version control system and suggesting learning opportunities immediately after a `git commit` is executed. To further assist in onboarding and codebase familiarity, the system includes the `orient` skill, which generates an `orientation.md` file to help the user navigate the project. The basic invocation is simple:

orient

Alternatively, developers can integrate Simon Willison's `showboat` for advanced code visualization and exploration. When the `orient` argument is passed to `learning-opportunities` in this context, the system generates two targeted lessons focusing on the core functionality of the repository.

The Case for Intentional Friction

Most AI tools are designed to remove friction. The goal is typically a seamless path from prompt to production, where the AI anticipates every need and provides the answer instantly. However, the core insight behind these new learning skills is that friction is exactly what the human brain requires to learn. Agentic coding, while efficient, often bypasses the struggle of problem-solving, which is the primary driver of long-term memory and expertise. By removing the struggle, these tools inadvertently accelerate the decay of the developer's own technical capabilities.

This repository introduces a deliberate break in the flow. When Claude triggers a learning session, it does not simply provide a summary of what it did. Instead, it asks a challenging question and then stops. It refuses to provide the answer, forcing the developer to pause and retrieve the information from their own memory or analyze the code to find the solution. This reversal of the standard AI interaction model transforms the AI from a ghostwriter into a tutor. It shifts the user's role from a passive reviewer to an active participant in the architectural process.

Recognizing that absolute rigidity would hinder productivity, the system includes safeguards to prevent learning prompts from becoming a nuisance. The tool is programmed to suppress learning suggestions during urgent hotfixes or when changes are trivial. This ensures a balance between the need for rapid deployment and the necessity of professional growth.

Beyond the immediate coding loop, the system integrates the `Learning-Goal` skill. This tool utilizes Mental Contrasting with Implementation Intentions (MCII), an evidence-based psychological technique where the user visualizes a goal and simultaneously identifies the obstacles that might prevent its achievement. This helps developers set interactive, realistic learning targets. To support team-wide adoption, GitHub has also provided a `MEASURE-THIS.md` playbook and a set of measurement tools released under the CC-BY-SA 4.0 license, allowing organizations to track how these learning interventions impact their engineering culture.

In an era where AI can generate a thousand lines of perfect code in seconds, the true competitive advantage for a developer is no longer the speed of generation, but the capacity for critical internalization.