The recent paper from the Anthropic Fellows program examines how AI impacts skill formation. It replicates the structure of the 2025 MIT study and involves participants learning the Python Trio library. Half used AI tools while the other half did not. The findings indicated that AI users did not complete tasks faster and scored lower on retention quizzes. However, the poor results can largely be attributed to a subset of participants who spent excessive time retyping AI-generated code instead of leveraging it efficiently. The study raises questions about the balance between speed and learning. While some AI users managed to work 25% faster without a significant drop in quiz performance, reliance on AI for coding can lead to a loss of foundational skills. Software engineers are primarily focused on delivering results rather than deepening their understanding of codebases, which might necessitate setting aside time for manual learning in the future. The author argues that moving faster could allow engineers to tackle more tasks, potentially offsetting the reduced learning per task. Further complicating the results is the choice of AI model, GPT-4o, which may not have been the best option for the study. This raises concerns about the reliability of the findings regarding AI's efficacy in skill development. The paper concludes that while AI use can hinder learning, it does not conclusively prove that it slows down productivity. Investigating the long-term effects of AI on skill acquisition and job performance remains essential as the technology evolves.