An analysis of GitHub pull requests reveals that large language models frequently use certain words like 'spine', 'gate', 'lane', 'proof', 'truth', and 'seam' at significantly higher rates than expected, with 'spine' showing a 20x increase in 2026 compared to 2025. The author uses GitHub Search data as a proxy to investigate this phenomenon, ruling out overall platform growth as the cause and noting that multiple LLMs including Claude appear responsible.