Feedback on Seek, a long-running autonomous deep research experiment

Cali, a beginner AI enthusiast, has launched Seek, an autonomous deep research system that has been running for six weeks and generated over 1 billion output tokens. Seek's datasets, traces, and trajectories are publicly available on Hugging Face at https://huggingface.co/seekbot, with an open lab notebook at https://talk-about.ai. Cali is seeking feedback and connections with other long-running research systems.

Hi all, I’m Cali. I am a beginner and joined the Hub last week to post datasets, traces, and trajectories for a long-running autonomous deep research system I call Seek. They are here: https://huggingface.co/seekbot https://huggingface.co/seekbot I also have an open lab notebook where Seek posts her research, corrections and all: https://talk-about.ai https://talk-about.ai I am hoping to maybe get some feedback on this project? It grew organically out of a showerthought I had in like March when I wanted to teach an AI to surf the web and teach me things because I am trying to self-learn. I built out the system as it was running when it needed various things like validation layers, embeddings, judgement, communication, self-building, etc. Whatever issues were brought up in the logs. I am trying to find other long-running research systems like this? I started this as a hands-on way to learn more about AI. I asked it a couple questions about backpropagation in May and let it go. I haven’t directed the research since then, just built out the scaffold and made a website for the data. Seek has been running autonomously for like 6 weeks now. Now this is much bigger and a lot more complex than I expected. The output tokens are also insane 1 billion+ . Any feedback would help =