Alan Yahya1 min read As far as we know, this is a novel combination of features to strengthen AI audit.
Down to each individual word. We attribute:
Whether the word was added by a human, or AI.
If AI, whether the suggestion was modified, or accepted as initially proposed. Also, which sources the AI cited, with specific pincites.
Also, the full trace of the agent conversation, including all tool calls and metadata (such as model involved, complete token composition, etc).
We trace the evolution of every paragraph in the document, from version 1, to the most recent version. For every version, you once again uncover the human or AI attribution with full agent traces.
Finally, if the edit was generated from a multi-agent swarm, we show the file lock, and any interactions that occurred between two agents that touched a paragraph.
You can see this in a quick video below.