Free tool
The Vibe Coding Cost Calculator
See how much of your AI coding budget is going to rework instead of shipping — and what it'd look like after a rescue.
Model based on Mockus & Weiss, “Predicting Risk of Software Changes” (Bell Labs Technical Journal, 2000), Stripe’s Developer Coefficient report, and GitClear’s 2026 AI Code Quality Research.
Quick start
01Total lines across your main repository, excluding lockfiles and generated assets. Used to project what a rescue could bring this down to.total, in your main repo
git ls-files | grep -vE '\.(lock|json|svg|png|jpg)$' | xargs wc -l | tail -102How many distinct domains, services, or table groups a typical change touches. The single strongest predictor of whether a change breaks something, per Bell Labs' analysis of 15,000 production changes.
Examples: a login feature touching only auth ≈ 1–2. A checkout flow touching billing, inventory, and notifications ≈ 3–5.
Examples: a login feature touching only auth ≈ 1–2. A checkout flow touching billing, inventory, and notifications ≈ 3–5.
03Roughly how many lines get added or modified for an average feature — not your total codebase size. Check the diff stat on your last few merged PRs.typical feature
04Bug fixes fail at higher rates than new features in production — the developer is usually working in code they didn't just write.
05How long the people actually shipping features have worked in this specific codebase — not total years of experience.months on this codebase
06How long your last real feature or fix actually took, start to finish — including review cycles and re-opened tickets. Not the estimate.
07A rough blended average across everyone who touches this codebase. $75–100 covers most small teams; higher for senior-heavy or contractor-heavy teams.
$
Model based on Mockus & Weiss, “Predicting Risk of Software Changes” (Bell Labs Technical Journal, 2000), an analysis of 15,000 production changes, and Stripe’s Developer Coefficient report. Context on why AI-era codebases trend this way from GitClear’s 2026 AI Code Quality Research (623M changes analyzed, 2023–2026): refactoring dropped from 21% to 3.8% of changed lines, duplication rose 81%.