The original title is "ThoughtDAG: Visualizing and auditing AI context compaction as a parallel graph"

Short summary

ThoughtDAG is a graph-based AI conversation tool that makes context compaction visible and auditable. Instead of silently summarizing history, it scans the conversation graph for collapsible runs, preserves human-marked decisions verbatim, and builds a parallel condensed tree alongside the original. Users can compare answers from both trees to verify the compression lost nothing important.

• •ThoughtDAG visualizes context compaction as a parallel graph rather than a black-box summary
• •Structural rules protect decisions, highlights, and the working frontier from being collapsed
• •Users can test compression fidelity by querying both the original and condensed trees simultaneously

Generated with AI, which can make mistakes.