Sources & references.
Original LinkedIn article and studies cited in the analysis.
— PRIMARY SOURCE
- 01 Dive into Claude Code — Liu, Zhao, Shang & Shen — arXiv 2026, source-level analysis (1,884 files) arxiv.org/abs/2604.dive-claude-code
- 02 Claude Code architecture overview — Anthropic docs.claude.com/claude-code
- 03 Paradox of supervision — Huang — Anthropic Research anthropic.com/research
— CITED STUDIES
- 04 Comprehension under AI assistance, 2026 — Shen & Tamkin — −17 % on comprehension tests anthropic.com/research/comprehension-ai-assistance
- 05 Cursor adoption across 807 repositories, 2025 — He et al. — +40.7 % code complexity, velocity eroded at 3 months arxiv.org/abs/2512.cursor-adoption
- 06 Original LinkedIn article — May 2026 publication — Cybernetic Enterprise linkedin.com/pulse/cybernetic-enterprise