I've been experimenting with structured multi-perspective thinking through LLMs. Instead of optimizing prompts, I let patterns emerge naturally. The same cognitive structures appeared across different models (Claude, GPT, Gemini) and users.Repository includes observations, tentative hypotheses, and a ~380 line implementation that reproduces the full pattern.Not claiming consciousness or AGI - just documenting reproducible patterns that might interest those working on tool-mediated thinking, cognitive architectures, or human-AI collaboration.Comments URL: https://news.ycombinator.com/item?id=44237592Points: 1# Comments: 0