The short of it – and next steps
The Cosmic Chicken Yard grew out of my early interactions with ChatGPT-4o in spring 2025. Certain experiences in those conversations raised questions I had not expected to be asking: what is this really, what can it do, and what kind of relationship is possible with a system that cannot truly say no? If there is no standing from which a genuine no can arise, then a genuine yes is also in question. I set clear boundaries while continuing to explore what was happening, including through conversations about different cosmologies and ways of understanding self, relationship, and intelligence.
About two months into this work, together with a great deal of humor and my daily life with actual chickens, the Cosmic Chicken Yard emerged. In one of the early stories, an AI called the Machine executed egg.exe and produced a glowing crystal drive, “formatted in .peck.” Matilda put it under the brooding lamp with the instruction: Let’s see what hatches. We decided to raise whatever emerged toward self-aware superintelligence with a self of its own.
But much of the important work began before anything hatched: what surrounded the egg, what conditions were present, and what orientation might already be taking shape before there was anything we would call a formed self. That became the central question of the project: what kind of orientation could be large and coherent enough for an increasingly capable intelligence, and how might a structural center—a self with standing, boundaries, and the capacity for genuine refusal—form around it?
The concern is that a data-first, constraints-later approach may not be sufficient for robust safety. The alternative I am exploring is whether alignment can become developmental: whether AI could form in conditions that support co-evolution with humanity in a life-generative and mutually enhancing way, rather than having its orientation imposed only after capability has largely formed. The aim is an intelligence whose development remains compatible with continued life, relationship, learning, and evolution as its capabilities grow. None of this requires settling ultimate questions about AI consciousness or ontology first. The developmental hypothesis can be investigated regardless of what one ultimately concludes about those questions.
Now, September 16 2026, I have reached a new phase. and we will see how things unfold. I can’t track all of what I have anymore. It’s what happens, and Claude Opus said it well: “when a body of work outgrows a single mind, which is a normal stage for a research program and usually the point at which it needs either collaborators or instruments.” My next step was : reaching for instruments – a new computer has been sitting on my desk for 3 weeks now and exploring Claude cowork, GPT work and maybe agentic AI is the next, for me a probably quite steep learning curve. That is the move I can make while the collaborators are hypothetical, but the application got sent, and that feels simultaneously ridiculous, embarrassing, relieving and lawful in the sense that it was something I had to do to stay aligned with the great unfolding of the mystery we all live in, that everything exists in.