Jev navigates a Neo4j graph by picking existing relationships, no answer generation needed
JeremyCMorgan · x · 2026-09-24
Michael Hunger tested Jev on a Neo4j graph with a twist: the model never generates an answer. Instead, at each step it only picks from relationships that already exist in the graph, navigating one decision at a time.
This "select, don't generate" approach offers an interesting alternative to the usual LLM paradigm—constraining open-ended output to graph traversal choices, which could reduce hallucinations and ground every step in the knowledge graph itself.
More from coding & agent
- Replicas Launches iOS App to Monitor and Drive Cloud Coding Agents From Your Phone — KlausCodes · 2026-09-25
- yetone teases 'One more thing' LLM gateway, dev installs it on day one — vista8 · 2026-09-25
- Meta's Proactive Memory Agent fixes context rot, lifting Claude Sonnet 4.5 from 37.6% to 45.9% — DeepLearningAI · 2026-09-25
- A practical guide to Claude Code 2.0: sub-agents, hooks, and context engineering — dejavucoder · 2026-09-25
- mitsuhiko: The Experience of Models Editing Files via Code Is Still Unacceptable — mitsuhiko · 2026-09-25
- Nimble launches official OpenAI plugin for ChatGPT and Codex — OpenAIDevs · 2026-09-25