Overview
- Anthropic publicly shared findings from an internal August study that surveyed 132 engineers and researchers and conducted 53 interviews about its agentic coding tool, Claude Code.
- Participants reported feeling more productive and more “full stack,” with the study estimating that 27% of Claude-assisted work comprised tasks that otherwise would not have been undertaken.
- Respondents said they could fully delegate roughly 0–20% of their tasks to Claude, particularly routine or easily verifiable work.
- Engineers cited reduced colleague interaction and fewer mentorship moments as AI became the first stop for questions, raising concerns about team learning and culture.
- Workers voiced fears about deep skill atrophy and long‑term job relevance, even as some embraced a shift to higher‑level abstraction and advised focusing on supervising AI and core concepts; external polling from McKinsey shows substantial interest in workplace AI tools.