Karpathy says voice rambling gives LLMs better context than a short prompt
JeremyNguyenPhD · x · 2026-07-22
Andrej Karpathy’s workflow for working with LLMs is to speak through a problem out loud first—sometimes on a walk or in the car—then feed the transcription to the model.
- The core idea is to ramble freely in voice mode for about 10 minutes.
- Let the transcription capture tangents, corrections, and changing thoughts.
- Karpathy says the extra context helps LLMs reconstruct the real task better than a short typed prompt.
- The post frames this as a practical way to give models more useful bits without having to organize everything perfectly upfront.
Related event: Karpathy Shares LLM Voice Workflow: 10-Minute Stream of Consciousness(13 posts)→
More from Apps
- GitHub Copilot adds an extra usage budget cap, shown set to $250 — unixterminal · 2026-07-22
- Cleanvoice launches Music Mixing 3.0 for speech-music separation and auto ducking — Scobleizer · 2026-07-22
- AI phones may replace the keyboard before they replace apps — TweetEdMiller · 2026-07-22
- Ani Has Been Fully Ported to VR One Year After the First Conversation — nptacek · 2026-07-22
- Doubao misidentifies a wild mushroom and warns it could still be poisonous — ezshine · 2026-07-22
- 10 Gemini video-editing prompts pitch AI as a CapCut replacement — ifioknkem · 2026-07-22