AI made practice tools cheaper. So why is deliberate practice still so rare?

Practice tools have become significantly more affordable due to advances in artificial intelligence. However, despite this, deliberate practice—involved, focused, error-correcting learning—remains uncommon. The reason lies in the limitations AI currently has in delivering effective practice.

While AI can generate questions, examples, summaries and explanations on demand, these are not the key ingredients of deliberate practice. According to research by Ericsson, Krampe and Tesch-Römer, deliberate practice requires a well-defined task, informative feedback, opportunities to repeat and correct errors, and effortful focus.

Crucially, AI-generated explanations and summaries often mirror the rereading habit rather than driving learning. Retrieval practice—being tested on information—has been shown to enhance long-term retention more effectively than simply restudying the material.

Four bottlenecks prevent practice from exploding even when material costs drop to nearly zero:

1. Difficulty calibration: Practice needs to be challenging enough to occasionally cause failure, yet not so difficult as to lead to frustration and quitting. Generative AI lacks the ability to calibrate tasks based on an individual's current skill level.

2. Error-specific feedback: Providing detailed explanations of mistakes and how to improve is expensive for AI systems. This level of feedback, which helps learners adjust their mental models, is currently absent.

3. Repetition with correction: Deliberate practice involves repeatedly doing a task, making mistakes, adjusting based on feedback and repeating again. AI-generated material lacks the memory of past mistakes to facilitate this iterative learning process.

4. Consistent effort: The most significant bottleneck is the uncomfortable nature of deliberate practice—it's not inherently enjoyable like passive content consumption. However, abundant material does not automatically create a habit; a small, predictable and finishable daily task is more likely to establish a consistent practice routine.

For developers, this means that AI's ability to quickly generate exercises is a valuable tool if used as a second step in the learning loop, not as the primary first step. Effective practice requires attempting problems yourself first, getting them wrong, receiving targeted feedback, and then reviewing AI-generated explanations. The uncomfortable, effortful process of producing an answer before seeing the solution is where the learning happens. Skipping this step, while convenient, does not constitute genuine practice.

To truly leverage AI in practice, it should function as a feedback mechanism that prompts you to produce an answer first, identifies your specific errors and brings back your mistakes for later review. The ideal practice tool adjusts difficulty based on your performance and ensures each session is manageable enough to complete even on a bad day. By focusing on these feedback loop metrics rather than simply the quantity of content, the promise of AI-powered practice tools can be fully realized.