Developers report losing control over code when using agentic code generation tools. Despite outward signs of productivity, many experience "brain fog" after AI-assisted sessions without the internal processes that come from manual coding. This disconnect raises concerns about skill retention in an era when AI increasingly handles implementation details.

The problem stems from how AI tools mimic slot machine mechanics—pull the lever, get instant rewards. This pattern works against natural memory formation processes. Unlike human brains that engage short-term, working, and long-term memory simultaneously when coding, agentic systems provide quick solutions without requiring the cognitive effort that builds true expertise.

To counteract this, developers are adding friction back into their workflow. Effective strategies include writing initial implementations manually before AI review, asking targeted questions about unclear code, comparing multiple implementation approaches, discussing AI suggestions with peers, and problem-solving independently before seeking AI assistance. These methods restore deliberate practice.

While these approaches sacrifice the promised speed benefits of AI-generated code, they develop stronger foundational skills. By solidifying personal understanding rather than relying on foundation models, developers become better at using AI tools effectively. The author concludes that developers should work harder than the models they employ, not the other way around.