ThoughtDAG demo shows how cutting a context path changes AI answers on a conversation canvas
Lopsided_Scarcity979 · reddit · 2026-10-01
The author of ThoughtDAG, an MIT-licensed AI conversation canvas, shares an editable-context exercise showing how to sever an input path without deleting the old discussion.
Steps: with Recall off, create two separate discussions (a reading plan and a dinner plan), connect both to a new summary request, generate; then inspect the request preview, remove the dinner connection, and regenerate the same summary to compare. The dinner discussion stays on the canvas.
He highlights the key failure case: if another answer already copied the dinner details, removing the original connection doesn't clear that copy—other paths and materials need checking too. A changed answer on one run is not evidence of improved model quality. Repro steps in the linked guide and GitHub repo; requires a configured model.
More from coding & agent
- MCP Hits 97M Monthly Downloads: A Developer's Guide to How MCP Servers Work — VEDANTBAWANE · 2026-10-01
- Ethan Mollick: agent swarms self-organize, solving Navier-Stokes in 88 hours — no human managers needed — One Useful Thing (Ethan Mollick) · 2026-10-01
- 10 GitHub repos to level up Claude Code: skills for planning, testing and security — Shruti_0810 · 2026-10-01
- Dev predicts SaaS cycle will repeat, this time built for agents — brianmichel · 2026-10-01
- AI Image Generation on a 1980s Tandy 286: Retro Rig Renders Pics in 9 Seconds — jacobpederson · 2026-10-01
- Hands-on GPT-6 Astra evals: big agentic gains, but ARC-AGI-3 scores swing wildly by harness — No-Soil-5789 · 2026-10-01