ThoughtDAG Lets You Edit the Hidden Graph Behind LLM Chats
Most chat interfaces for large language models treat a conversation as a straight line: each message follows the last, and if you want to explore a different direction, you either edit and resend or start a new thread. ThoughtDAG, a small open-source project shared on Hacker News, reframes this as a directed acyclic graph (DAG) that users can branch, prune, and rearrange directly.
The idea is that context is not really linear in practice. People backtrack, try alternate prompts, merge ideas from different branches, or want to reuse a chunk of context in a new direction. By exposing the underlying structure as a graph you can manipulate, ThoughtDAG aims to make that kind of nonlinear exploration a first-class feature rather than a workaround.
It's an early, low-traffic Show HN project rather than a polished product, but it points to a broader trend of rethinking chat UI as LLMs get used for more complex, multi-step reasoning tasks.