Mace-legal teaches Mike to read redlines

Until now, Mike read a marked-up document the same way it read a clean one - it saw the words, but not who struck what, who added what, or what a reviewer scrawled in the margin. Mace-legal closes that gap. The fork pulls insertions, deletions, and comments out of Word documents, and reads color-coded redlines out of PDFs in the Litera and Workshare style - the redlining tools most law firms use to negotiate contracts. Every marked-up document gets translated into a consistent shorthand the AI is now taught to interpret.

Crucially, the model is told how to behave: treat the current text as if every edit were accepted, treat comments as marginalia rather than instructions, and keep markers out of extracted values unless you actually ask for them. Clean documents are untouched and pay no penalty.