26
Control participants (No AI)
Main study group sizeQuick read of downside signals from Anthropic's 2026 paper on coding skill formation ("How AI Impacts Skill Formation").
Control participants (No AI)
Main study group sizeMedian total errors per participant
Q1-Q3: 2-5 errorsFinished Task 2 within 35 min
4 of 26 did not finish Task 2 in timeAverage quiz score vs AI group
Higher across all coding-experience bands (Figure 7)These cards are not AI-vs-No-AI. Every row below is from participants who had AI assistance; it shows which AI usage style performed better or worse.
More debugging-query reliance -> lower quiz score
Correlation r = -0.41 (p = 0.043)More debugging-query reliance -> slower completion
Correlation r = 0.43 (p = 0.033)