HomeAll NewsEducationTV
Equities & FundsCrypto & Digital AssetsAI & TechnologyBusiness & CorporateUS Politics & PolicyGeopolitics & Global RiskMacro, Rates & FXCommodities & EnergyEuropean Politics & MarketsAsia-PacificReal Estate & Property
Story archiveAll categories
← All Stories

Andrej Karpathy suggests rambling for 10 minutes to improve AI chatbot responses

Created at 24 Jul · 9:51 AM1 source↑ Market-relevant
IN SHORT

AI researcher Andrej Karpathy proposes a method for users to achieve better results from large language models: engaging in a 10-minute stream-of-consciousness monologue. This approach aims to provide the AI with more context, allowing it to better understand user intent and reduce the need for corrections.

✉Newsletter

PiQ Daily

Pick your topics. Get only what matters, on your cadence.

Key Numbers

10 minuteslength of AI interaction

Who's Involved

Andrej Karpathy
AI researcher at Anthropic and proponent of rambling prompts
OpenAI
Company developing AI voice tools like GPT-Live
Anthropic
Company that updated its voice models
Andrej Karpathy suggests rambling for 10 minutes to improve AI chatbot responses

↳ Why This Matters

This approach offers a potential method for users to improve their interactions with AI chatbots, especially as voice-based AI tools become more prevalent. It highlights a strategy for managing the complexity of AI prompting by offloading raw thought processes to the machine for organization.

Key facts

  • AI researcher Andrej Karpathy advises users to engage in lengthy, rambling voice prompts with AI chatbots.
  • Karpathy suggests this 'stream of consciousness' approach can help large language models (LLMs) better understand user goals.
  • The method involves speaking to the AI for approximately 10 minutes, unloading half-formed thoughts.
  • Karpathy claims this improves the human-AI connection and minimizes subsequent corrections.
  • This advice coincides with advancements in AI voice technology from companies like OpenAI and Anthropic.

Andrej Karpathy, a prominent AI researcher, has proposed a novel method for users seeking better responses from artificial intelligence: simply talk to the AI for an extended period. Karpathy, who coined the term 'vibe-coding,' suggested in a recent social media post that users should engage in a "total mess, anything goes, full stream of consciousness" monologue for about 10 minutes.

Karpathy explained that this approach provides the large language model (LLM) with more 'bits' of information, thereby enhancing its understanding of the user's objectives. He described the process as a way to improve the 'mind meld' between the human and the AI, ultimately leading to fewer necessary corrections. The AI researcher often begins by warning the chatbot about potential jumbled thoughts and typos in his input.

While Karpathy advocates for this verbose interaction style, some users on the platform X have expressed skepticism. His advice emerges at a time when major AI companies are actively developing and promoting voice-enabled AI tools. OpenAI recently introduced GPT-Live, which powers ChatGPT's voice capabilities, and Anthropic has updated its own voice models, offering users choices between its Opus, Sonnet, and Haiku offerings.

Frequently asked questions

Andrej Karpathy suggests users engage in a 10-minute stream-of-consciousness monologue with the AI chatbot.

He believes it provides the AI with more context ('bits') to better understand the user's goals and reduces the need for subsequent corrections.

Vibe-coding is a term coined by Andrej Karpathy, likely referring to the intuitive or less structured way of interacting with AI.

Yes, OpenAI has GPT-Live for ChatGPT Voice, and Anthropic has updated its voice models.

What Happens Next

01Users may experiment with longer, rambling prompts to test AI responses.
02AI companies may further refine voice input and processing capabilities.

Get the newsletter.

Pick the topics you actually care about. We'll email when there's news worth your time, on the cadence you choose. Cancel any time from your account.

Cadence
CME Headlines
  • Is AI Making Inflation Better or Worse?
    22 Jul · 3:26 PM

How It Developed

Andrej Karpathy suggested users ramble at AI chatbots for about 10 minutes.
Karpathy stated this approach helps LLMs understand user intent by providing more context.
He noted that this method improves the 'mind meld' between human and AI, reducing corrections.
Other users expressed skepticism about Karpathy's suggestion.
The advice comes as companies like OpenAI and Anthropic are enhancing their AI voice tools.
Sponsored

London Quick Take - 22 July - UK inflation softens, oil rises and chips rally ahead of Alphabet, Tesla earnings

SAXO

Sources

T1
Andrej Karpathy’s trick for getting better answers from AI: Just ramble at it for 10 minutesBusiness Insider

Related Stories

Musk proposes AI safety peer reviews, reflects on political involvement
23 Jul · 11:45 AM
Black Forest Labs Launches FLUX 3 AI for Video Generation, Robotics
23 Jul · 10:06 PM
Anthropic economist: AI won't replace workers; Cuban highlights human advantages
23 Jul · 12:36 PM
US Officials Accuse Chinese AI Firm of Stealing Tech; Experts Skeptical
23 Jul · 11:16 AM
SAP CFO: AI must evolve beyond chatbots for real business returns
23 Jul · 9:55 PM