Us-vs-them Tool Scores Code Lines by Author

A Clojure tool called us-vs-them scores each code line 1.0 for human, 0.0 for agent, using only git history, per its GitHub repo. That gives anyone running an AI coding agent a way to flag hand-written sections so they don't get silently rewritten. It doesn't touch the file or add a watermark.

The scores aren't binary. A line sitting at 0.46 means a human wrote it first and an agent modified it later. The tool tracks how blocks of code join, split, and dilute across commits. It's built to find coherent chunks of authorship, not just tag lines one by one.

The stated use case is --ours/--theirs protection: mark a section as yours, and the tool catches an agent trying to overwrite it. It's picked up 51 points on Show HN and sits at 41 stars. That's modest, for a tool answering a question more developers are asking as agents write more of the diff.

The catch is that the whole system runs on commit history staying honest. A squash merge, or an agent committing under a human's name, would erase the signal the scoring depends on. The repo doesn't say how it handles either case, which matters more as agent commits start looking indistinguishable from human ones.