Blog
How we built Source Trace, and what the data says about AI-generated code.
-
What code rework means for a small team
AI coding spend adds up fast on a small team. Source Trace for Teams gives you a clear answer on which tools are worth paying for: adoption, kept code, and excess rework.
-
Survival rate for CLI-first agents
More code is written by terminal agents that never open an editor. Source Trace now attributes and scores CLI-first tools the same way it does IDE assistants.
-
Beyond co-authored-by: attribution and ROI across vendors
A single 'Co-authored-by: Copilot' trailer can't tell you which AI tool earns its licence. Per-model attribution turns multi-vendor spend into measurable ROI.
-
Opening the public beta
Source Trace is now open to everyone, free. One-click install, nothing to configure, and your source code never leaves your machine.
-
Survival rate: what you keep, not what the agent wrote
AI agents write code straight into your files. Survival rate measures how much of it is still there at commit: the pre-commit gap that matters.
-
Zero-config by design: passive collection for VS Code
If measuring AI code costs you a workflow change, you won't do it. Source Trace installs in one click and collects attribution passively, with no config and no new habits.
-
The measurement gap in AI-generated code
Teams know how many tokens they burned, not how much AI code they kept. We built Source Trace to close that gap. Today we're opening a private beta.