I built an AI interface that refuses to give advice

Over the past year, I’ve been thinking about how AI interfaces are evolving.

After watching Perfect Days, I became interested in the idea of slowness and attention in everyday life — and whether software could reflect that instead of constantly optimizing for speed.

Around the same time, I saw Andrej Karpathy’s “LLM Council” idea — multiple AI voices responding to the same question — and wondered what would happen if that idea was applied to reflection rather than intelligence.

So I built a small weekend project called The Hirayama Table.

You bring a question to the table, and five philosophical voices respond — Lao Tzu, Krishnamurti, Gurdjieff, Osho, and Khalil Gibran.

They don’t recommend. They don’t solve. They don’t converge. They reflect.

It’s less a product and more an interaction-design experiment: can AI act as a mirror instead of an answer engine?

Would love feedback from this community.