Karpathy says long voice rambles can improve LLM task framing and RAG use
anshulkundaje · x · 2026-07-22
The post highlights a workflow where long, messy voice rambles can be useful input for LLMs, especially when combined with a RAG database or knowledge base.
- Karpathy’s point: speaking freely for several minutes can help the model reconstruct intent from an incoherent train of thought.
- The added comment argues that auto-injecting such rambles into a RAG vector DB or knowledge base can be beneficial even when the match is imperfect.
- The practical takeaway is that LLMs can turn raw, messy thinking into cleaner task framing and better downstream retrieval.
Related event: Karpathy Recommends 10-Minute Voice Rambling for LLM Context(17 posts)→
More from coding & agent
- TRL adds Score Centering, a simple fix for RLHF training instability — lvwerra · 2026-10-27
- Precompute golden answers and match at runtime: an LLM pattern from AustinLangChain — colinmcnamara · 2026-10-27
- Precompute golden answers and match at runtime for exact wording control — colinmcnamara · 2026-10-27
- Dev argues there's still alpha in working with your coding agent interactively — alex_frantic · 2026-10-27
- Researchers launch AMA on patterns for safer AI agents used by citizen developers — Loose_Rip359 · 2026-10-27
- Gradio Adds In-Chat Human-in-the-Loop So ML Intern Pauses Before Training — Gradio · 2026-10-27