LLM Chat with Open Threads

The seed

Most chat interfaces treat a conversation as a single line: one message after another, one topic until the next, and when you close the window the thread is effectively gone. Real thinking does not work this way. It forks — one question opens three — and the forks loop back, informing each other days later.

This quest is a chat experience built around open threads: conversations that can branch into parallel lines of inquiry, reconnect when two lines turn out to be the same question, and persist so that a thread you started last month is still there, still live, ready to be picked back up rather than re-explained from scratch.

Why it matters

A conversation with a model is a form of thinking, and thinking deserves a substrate that remembers. When threads persist and interlink, the chat stops being a disposable transcript and becomes a knowledge structure you can return to, navigate, and build on — the same regenerative, accreting quality EvoBioSys wants in all its tools.

• Branching — explore several directions from one point without losing the others.
• Reconnection — merge threads that converge, so insight isn’t stranded in a dead tab.
• Persistence — conversations survive the session and stay reachable across time.

What building it out looks like

• Model the conversation as a graph of branching, rejoinable threads rather than a linear log.
• Design an interface that makes forking and reconnecting threads a natural, low-friction gesture.
• Persist threads durably and locally, so a session left open weeks ago resumes exactly where it stopped.
• Keep the model layer pluggable so the experience can run against local or self-hosted inference.

Status

Dormant. The concept is captured but no implementation has begun. It awaits a decision on the underlying chat and inference stack before any building starts.