Data-driven studies of AI coding assistants (Claude Opus 4.7, Codex, Cursor, etc.)
31 articleWe counted every shadcn/ui component import across 12K Opus 4.7 repos. Four components dominate — the rest are long tail.
When Opus 4.7 writes Python, 84% of functions declare return types. That's 2× the typical community rate.
Median file size by language across the Opus 4.7 corpus reveals a clear ordering — and it matches the language's native verbosity.
The correct pattern (.env.example) exists in 1,877 repos. The leaky pattern (.env in git) exists in 52. Small number, real problem.
Every AI coding tool has its self-documentation file. Across the Opus 4.7 corpus, CLAUDE.md appears 50× more often than .cursorrules.
12,095 repos, 84 GitHub Actions workflows. That's under 1%. Here's the full story of Opus 4.7's automation gap.
Counting actual page.tsx files vs pages/ files across 12K Opus 4.7 repos. App Router wins 2:1.
3,006 TypeScript repos. 3,006 tsconfig.json files. No "JavaScript wearing a costume" in the Opus 4.7 corpus.
Despite tRPC's popularity in "modern TypeScript" discussions, Opus 4.7 uses it in only 17 of 12,095 repos. Here's why.
When Opus 4.7 needs a backend-as-a-service, Supabase wins by 3.5× over Firebase. The numbers explain why.
Which framework combinations actually appear together? The numbers reveal clear archetype patterns.
The median Opus 4.7 README is 66 lines. The p90 is 267. The maximum is 6,827. This is genuine documentation discipline.
Our research is based on continuous analysis of 128,000+ repositories and 3.27 billion lines of code using Repobility's proprietary scanning engine.
All data is aggregated and anonymized. No individual repository names or source code is disclosed.
Access our proprietary datasets for your own research, product development, or competitive intelligence.
Browse DatasetsGet our latest research and intelligence reports delivered to your inbox.
No spam. Unsubscribe anytime.