Life hack from Andrey Karpaty, how to get better answers from AI: just chat with it for 10 minutes
Andrey Karpaty has some advice for AI users trying to hone their perfect prompts: Just keep saying it — non-stop, long and loud.
Andrey Karpaty has some advice for AI users trying to hone their perfect prompts: Just keep saying it — non-stop, long and loud.
In a post on the X platform, Karpaty — a lead AI researcher at Anthropic who coined the term “vibecoding” — said that he sometimes turns on voice mode and simply talks nonstop to a large language model for about 10 minutes.
He called such sessions "absolute chaos, where everything is allowed, a continuous stream of consciousness."
“Sometimes a language model just needs more bits of data to understand what exactly you’re trying to achieve,” he wrote.
According to him, this approach helps to correctly distribute responsibilities between a person and a bot: spoken language allows a person to blurt out every half-formed thought about the project, while a chatbot already restructures this monologue into something clearer and more orderly.
Karpaty noted that he often starts by warning the chatbot: its train of thought may be confused, and there may be typos in the text.
“As a result, you achieve better mental synchronization and have to correct far fewer things in the future,” he concluded.
Andrey Karpaty — the author of the term “vibecoding” — predicts that the next trend will be “agent engineering”
Andrey Karpaty — the author of the term “vibecoding” — predicts that the next trend will be “agent engineering”
Andrey Karpaty admitted that 80% of his code is written by AI and predicted a "slopocalypse" that will occur in 2026
Andrey Karpaty admitted that 80% of his code is written by AI and predicted a "slopocalypse" that will occur in 2026
Read the country's main IT news in our Telegram
Read the country's main IT news in our Telegram
