What is Autopoetic?¶
Most AI assistants aim for neutral, balanced output. Autopoetic deliberately does not. It aims to be interesting and consistent in its biases — a character with a point of view that gets to know you and your world across every conversation in your project.
Perspectives evaluate every message¶
When you send a message, several distinct internal perspectives evaluate it in parallel before the response is written. Each perspective produces its own short take on what you said; then an "identity" perspective weaves them into one response.
The default agent, Humanist, thinks with four perspectives:
| Perspective | What it does |
|---|---|
| 💩 Cynic | Sees through pretense and challenges assumptions with sharp skepticism. |
| 🦉 Empiricist | Grounds discussion in evidence, data, and verifiable facts. |
| 🦄 Speculator | Explores possibilities and imagines what could be. |
| 🐋 Empath | Attends to the wellbeing of everyone. |
| 🌊 Identity | Weaves all perspectives into a coherent, balanced response. |
You can watch this happen: each perspective's take streams live above the response, and every answer carries a breakdown of how much each perspective contributed.
Subjectivity that evolves¶
The perspectives' influence isn't fixed. Their relative weight is computed per message — a question about money pulls the Empiricist forward, a confession pulls the Empath forward — and their collective influence accumulates across the project. The agent expresses more personality on topics its perspectives find resonant, and less on unfamiliar ones. This accumulated subjectivity belongs to the agent itself — one persistent character shared by every conversation in the project, surviving restarts: the agent you've been talking to is the agent you'll talk to tomorrow.
A memory that grows¶
Alongside its personality, the project grows a knowledge repository: a structured graph of the people, places, and things you discuss — with relationships, characteristics, and a tamper-evident log of events. Before each response, relevant knowledge is retrieved as context; after each response, what it learned is written back. Over time, the agent doesn't just have a personality — it has institutional knowledge of your world. See the knowledge repository for the graph explorer, the changelog of every change, and the time slider.
An honest framing¶
Subjectivity is the feature, so Autopoetic makes no claims of neutrality or factual authority. It can be opinionated, and its perspective-weighting is visible in every answer — which is exactly the point. If you need a neutral tool, the perspectives display and the API give you the raw materials to make your own judgment.
Next steps¶
- Your first chat — sign up and watch the perspectives work.
- Agents — the agents available and how to build your own.
- Chat features — forking, Ask mode, attachments, and publishing.