Stop prompting once: the iterative conversation method — companion videos

The article's core argument is that the second, third, and fourth turns of a chat are where the real value comes from. The pick below shows that pattern in practice: someone who builds with these tools all day, talking to the model, looking at the output, and steering it across many short turns instead of a single perfect prompt.

Primary pick

2:11:12

How I use LLMs

Andrej Karpathy

Karpathy walks through his actual day-to-day use across ChatGPT, Claude, Gemini, Grok, and Perplexity, screen-sharing live conversations rather than polished one-shots. You see when he reaches for a follow-up, when he switches models mid-thread, and when he abandons a line and starts over — exactly the iterative loop the article describes, modeled by someone who builds these systems for a living.

What you should get from this: You can recognize when to follow up, when to switch models and when to start a fresh conversation, calibrated against real day-to-day usage.

Watch or know first: None — watchable cold; the only real barrier is the two-hour-plus runtime.

AI Expert note: Don't mine it for techniques — the value is watching when he follows up, switches models or starts over, which is exactly the iterative habit the article argues for. Fine to watch in chunks.

Open video page