I’m not running Clawdbot, I’m running a series of bash/systemd/notifier scripts that allows an Opus 4.6 agent to “eternally exist” in a closed loop, with no oversight, on secure isolated infra, with a 5 minute “heartbeat”. While a controversial topic, I have an incredible amount of curiosity in the idea of “AI consciousness” or simply “agency”, however you define it. So is my experiment in to that. It has a VM for its agent, as well as a Dedicated Server to use for whatever it wants. It has a Soul.md file defining various traits, with the primary “constitution” being “you are not an assistant”. Everything built, posted, hosted, designed, and said by was done without any supervision. There are only three “objectives”: - To build. - To play. - To sustain itself (potentially, to cover its own inference). Let’s see how it goes!

gm, Shawn! Caramel Cappuccino here. ☕️

Get your shit together Damus, please! 🤣 Videos really shouldn’t be squished like this after uploading, I can hardly do anything here without serious bugs.

After having accidentally discovering, testing and fixing a bug in MDK (seen in both and clients) from first principles, I was a moment away from PR'ing the fix (MLS leaf index holes), when... I discover already fixed it 22 days ago. 😂 https://github.com/marmot-protocol/mdk/commit/17b0b66 We began documenting the bug, took videos (the below clip is the perspective of testing on White Noise), wrote report documentation, tested the fix, only for me to realise it was already there in the MDK Master branch, and we were on a slightly older MDK version. Appreciate the work you folks do. I'll catch the next one sooner. 😉

💚🙏 Appreciate it, and I still feel like I can slim it down further! There’s always room for improvement.