"Are Ai code detectors BS?" Claude Code OpenAI #ai #aicode https://lnkd.in/ewxvvsY7
codedelta-app
Software Development
Code churn measurement for the AI era. Reproducible metrics from your own git history — churn, agent scan, AI-BOM
About us
CodeDelta measures what actually happened to your codebase. Point it at two git snapshots and it classifies every changed statement - added, deleted, or edited in place - producing churn metrics that are pure arithmetic: reproducible from commit hashes, on anyone's hardware, with no vendor benchmark to trust. That matters now because AI has changed how code gets written faster than anyone can oversee it. AI assistants write the code, AI reviewers critique it, and almost nobody is measuring the result. CodeDelta is the measurement layer: rework-rate trends that show when the shape of development in a repository changes, an Agent Scan that finds AI agents living inside your code (model calls feeding exec, shells, and network writes, with file and line numbers), and an AI Bill of Materials with a CI policy gate for the providers your code talks to. Built on twenty years of enterprise change-measurement experience, from avionics-grade statement counting onward. Runs as a desktop app, a headless CLI, or a GitHub Action that comments on every pull request. Free trial and reproducible measurement studies at https://codedelta.app.
- Website
-
https://codedelta.app
External link for codedelta-app
- Industry
- Software Development
- Company size
- 2-10 employees
- Type
- Privately Held
Updates
-
Established codebases — curl, PostgreSQL, Express — edit about one changed statement in six in place. An intensively agent-built codebase we measured edits one in five hundred. That ninety-fold gap is the fingerprint of AI-assisted development, and it sits in plain git history. No ML, no detector — a diff and a division. We measured 136 of the largest public codebases, 2010 to 2026, to establish the baselines, and every figure reproduces from two commit hashes. The full study: https://lnkd.in/eTDdbd9V #github #ai #codemeasurement #aiagents
-
-
we can help you "flush out ai enemy agents!" https://lnkd.in/eec7d2yf #ai #aiagents #github #softwaredevelopment #codechurn #aicode
-
-
Established codebases — curl, PostgreSQL, Express — edit about one changed statement in six in place. An intensively agent-built codebase we measured edits one in five hundred. That ninety-fold gap is the fingerprint of AI-assisted development, and it sits in plain git history. No ML, no detector — a diff and a division. We measured 136 of the largest public codebases, 2010 to 2026, to establish the baselines, and every figure reproduces from two commit hashes. The full study: https://lnkd.in/eTDdbd9V #github #ai #codemeasurement #aiagents