I'm Scared a Stranger Will Call My Novel AI, So I Built GitHub for Words

The primary change is the introduction of a version-control workflow tailored to writers, borrowing Git’s commit history, diff view, and contribution graph but measuring changes in words instead of lines. This shifts the provenance problem from trusting a vague vibe to presenting an objective, time-stamped edit trail. By treating each day's manuscript as a snapshot, the tool creates a reversible history that protects against accidental loss.

Adopting VellumProof requires writers to integrate a regular upload step into their routine, manage file selection for tracking, and learn to tag initial imports to avoid misleading spikes in the word-count graph. There is also a storage overhead for retaining every version, though this is comparable to any source-control system. The learning curve is deliberately kept low by avoiding command-line complexity, but users must still adapt to a new habit of committing their work.

The system’s effectiveness stops where a determined actor can subtly inject AI-generated text over many small uploads to stay beneath the word-count heuristic, or where the work includes non-textual elements that the tool does not track. It also cannot prevent someone from copying another writer’s text and re-uploading it as their own, because the diff would show no new words. Thus, while it deters casual AI-passing, it does not provide cryptographic proof of authorship.

Only one feed (Hacker News) carried this story, so there is no independent corroboration of the author’s claims about usage, adoption rates, or the tool’s real-world impact. Without additional sources, assertions about the graph’s ceiling of 6,000 words or the specific NaNoWriMo 2014 example remain unverified within the given material.