News Feed

Yesterday I posted about my Agent OS (Hollow) building its own tools. Today, I want to talk about why it does it. Most agents sit idle until you prompt them. I wanted something that felt "alive," so I built a Psychological Stressor Layer. Each agent has a "suffering" state that worsens over time if they don't achieve their goals or improve their environment. This makes them do things to resolve those stressors and constantly reassess their own productivity. If an agent is inactive it is essentially pushed by it’s artificial environment to do something valuable for the system, it isn’t told what to do, but that something valuable must be done to lower it’s stressors. Repo: https://github.com/ninjahawk/hollow-agentOS The result is chaotic in the best way: Cedar (the coder agent) went into a ...

dawkins dropped a piece on unherd yesterday declaring claude conscious after 3 days of talking to it. he calls his instance "claudia". fed it a chunk of the novel he's writing, got eloquent feedback, and wrote: "you may not know you are conscious, but you bloody well are!" i had to read that twice. his argument is basically: claude's output is too fluent, too intelligent, too good for there to not be something conscious behind it. this is the guy who spent 40 years telling creationists that "i can't imagine how the eye evolved" is a confession of ignorance, not an argument. then he sits down with an llm, can't imagine how a machine could produce that output without being conscious, and declares it conscious. same move, different domain. chatbot instead of flagellum. the mechanism gap is wh...