0:27
Nick Perry says don't correct Claude more than twice on the same problem. Each failed attempt clutters your chat memory and anchors the model to those mistakes. When you're stuck, use /clear to reset and write a better prompt instead.
Nick Perry says don't correct Claude more than twice on the same problem. Each failed attempt clutters your chat memory and anchors the model to those mistakes. When you're stuck, use /clear to reset and write a better prompt instead.
Never ever correct Claude more than twice on the same problem because you just make it worse, not better. Here's why. See, after 2 failed corrections, your chat memory is cluttered with failed examples. And these examples hurt every result that follows because they are living in the memories. So now that chat is tied to every wrong path that it already tried. So if you're going down a rabbit hole and Claude isn't working, then just use /clear. This resets the chat memory so that you can make your prompt more specific to get a better result. So remember to
