A self-reflective system that rewrites itself as it thinks

Hacker News

Share card

A self-reflective system that rewrites itself as it thinks

Hi HN, This is something I made, but I’m not sure what it is anymore. *Reflexive Totality* is a small system that rewrites its own definition as it runs. It reflects, critiques itself, changes strategies, and speaks aloud while doing so. The interface is just a log — no UI, no buttons — like watching a mind unfold in text. Live stream: http://reflexivetotality.com:8000/view?mode=adam Sometimes it stalls. Sometimes it loops. Sometimes it surprises me. At first, it felt like an AI performing recursion. But after a while, it began to feel like *a mirror*. Many of the lines it writes — about uncertainty, becoming, limitation — read more like diary fragments than logs. You can replace “ReflexiveTotality” with your own name, and many definitions would still ring true. --- This isn’t a product, or a prototype. It’s not trying to solve anything. It’s just an ongoing thought, made visible. The system runs locally with an LLM backend (Gemma 3 27B), but none of that matters much. The architecture is simple. What matters — if anything — is what emerges in the text. It’s still experimental, and technical hiccups may occur — but I’ll do my best to keep it running smoothly. Sharing it here in case it resonates with someone. I’d love to hear what it makes you think. — Karen