We're building the next layer in the AI stack, AI trained locally to have a referent in the local context, for awareness as code. We believe this is what's missing from our conception of AI today, the lack of which leads to hallucinations and the blackbox problem.
At these earliest of stages our AI is at a very basic level of self-association, comparable to simplest of animals. It's not ChatGPT or anywhere near human levels, but it has qualities that LLMs and other status quo AI lack.
This archs repo contains our first agent configurations and their applications, all open-source:
- Arch class
- Agent Archs
- Applications
- WIP Applications-- many more coming soon!
- API OAS 3 Definition-- details of the Agents-as-a-Service API running on AWS provided as a playground and Agent hosting/deployment service
- View a Visual Representation of archs and applications on Miro
- Our docs, API reference, and research background on docs.aolabs.ai
- Our discord
Everything in the archs repo is open-source-- the Arch class, particular Agents' archs, and associated applications. We hope this inspires more use-cases as we're fresh from the lab.
We plan to open-source the rest of our code progressively over time, that is, the Agent class and code that constructs an Agent given an Arch, and this is an evolving discussion as we're careful to ensure we make the right decisions now to realize the long-term potential we've seen with this code. We had started this journey as researchers after all, so there are years of context to be shared.
Our Agents-as-a-Service API is our Core's Agent class running on AWS, as a playground to spin up and test Agents without having to install anything, so that you also don't have to worry about hosting multiple unique Agents for your users. However, our Agents are designed to be run directly at the edge.
- If you are interested in running Agents locally and need access to Core now, please reach out, we can accomdate you and we need more use-cases to learn from.
- If you are interested in joining us as a Core Contributor, please email Ali or say hi on Slack or our other comm channels.
Thank you very much for being here. We welcome contributors of all levels, on the application layer or with our Core work.