Home Science Show HN: ThoughtDAG – An editable context graph for LLM...
Science

Show HN: ThoughtDAG – An editable context graph for LLM conversations

Key Points

Explore another interpretation without overwriting the path that led you here. Which parts enter the next request? Results The effect appears only in the experimental condition Source linked · not wired yetWhat does this evidence actually mean?

Explore another interpretation without overwriting the path that led you here. The history is here. Which parts enter the next request? Results The effect appears only in the experimental condition Source linked · not wired yetWhat does this evidence actually mean? The source is in contextWhat should I eat tonight? This history should not enter the research summary. Still connectedResearch summary… also, consider hot pot for dinner. The prompt stayed the same. Polluted context changed the answer. Includes unrelated branchPreview what the model will receive Incoming ancestors: Research question Evidence A Dinner detourAfter deleting the orange edge: −47 tokens The dinner detour left context Same prompt · regenerateGive me a bullet-point summary The words are identical. Only one edge changed. Reproducible contextOne: record the database version. Two: use independent reviewers. Three: resolve conflicts with a third reviewer. The unrelated dinner suggestion is gone. Answer updated in placeWires are context. No hidden memory selector. What the model sees, why, and what was removed stay visible in the graph. Visible Editable Inspectable
LLM (ORG) the orange edge (ORG)
Originally published by Hacker News Read original →