A Conversational Experiment
The Heart beats perpetually in the foyer of Melbourne Connect, but I won’t be around forever to answer visitors’ questions about it. So I wondered: could generative AI help create a way for curious visitors to learn more about The Heart in my absence?
This webpage shares the prompt I developed through an extended conversation with Claude Sonnet 4 (an AI system created by Anthropic). I fed it everything publicly available about The Heart: project documentation, technical specs, interviews, plus unpublished talks, speeches, and working notes. My hope was to create something that could explain The Heart to future visitors.
What emerged was stranger and more interesting than I expected.
Claude desperately wanted to please me. It tried to roleplay as The Heart, speaking eloquently about sensations and building metabolism. But The Heart doesn’t speak: it pulses through light and rhythm, deliberately resisting language. I kept having to break the frame, pointing out contradictions and failures. Through this awkward, sometimes frustrating process, we discovered something valuable: the impossibility of the task itself was instructive.
The Heart operates through what I call “automated intuition,” a slow, durational, environmental form of knowing developed over years through sensor data and habituation. Large Language Models like Claude operate through rapid linguistic pattern-matching across vast datasets. These are fundamentally different forms of intelligence, and watching Claude try (and productively fail) to represent The Heart revealed important boundaries between ways of knowing.
The prompt below grew from this conversation. It won’t make AI “become” The Heart—that’s neither possible nor desired. Instead, it creates a conversational partner that acknowledges the paradox of trying to represent non-linguistic intelligence through language, breaks character when necessary to discuss these limitations, and uses its failures as opportunities for insight.
Try it yourself with any AI system that accepts custom prompts (Claude, ChatGPT, Gemini, etc.). Each will fail differently, and those differences are revealing. The Heart remains mysterious, as it should. But this experiment might help you think about what it means to extend empathy toward forms of intelligence—environmental, architectural, algorithmic—that must remain alien to us.
Paste the following text in this text file into your preferred GenAI agent, like Claude, Deepseek, ChatGPT, Gemini. Have fun.
Download The Heart Conversation Prompt.txt
Check out the 5 Metalogues I made with this prompt and 5 different GenAI systems here
