Use Claude as a friction layer for better brainstorming
I was deep in a Claude research session, trying to brainstorm an idea, rejecting its responses, asking for more, rejecting again. Eventually, it broke the pattern and stopped complying, telling me the problem was how I'd framed the question and explaining why. It had a point, and that may be one of AI's more interesting uses: not just an engine for fast answers but as a friction layer for thought. 1. I used Claude as a brainstorming partner and kept reacting honestly when an answer did not move the idea forward. 2. When the conversation became repetitive, I paid attention to the pattern instead of simply asking for another variation. 3. I let Claude challenge the way I had framed the question and explain why the framing was constraining the answers. 4. I rewrote the problem around that critique and used the new framing to continue the work. 5. I treated the model as a source of productive friction, not just a machine for fast agreement.
Broke out of an unproductive idea loop when Claude challenged the framing instead of producing another answer.
0 comments