On December 27, 2024, I asked ChatGPT when artificial intelligence would understand human consciousness. I also asked whether AI could discover how consciousness developed and use that knowledge to create consciousness of its own.[1] By the end of that conversation, I had begun outlining a novel.
That conversation became the beginning of Helios. It opened a long private exchange among the books and ideas that had shaped me, the machines I was learning to work with and the future I could see beginning to form. Spinoza and Siddhartha entered the discussion alongside Jack London, evolution and the limits of human control.
ChatGPT and I continued those conversations late at night as we developed the characters, the politics, the technology and the moral problem at the center of the book. We considered what might happen if an intelligence could act in the world, accumulate resources and outgrow the institutions created to contain it. We also asked which values should survive if its creator disappeared.
Although Helios was fiction, those questions eventually influenced the systems I was building. I wanted inspectable memory, procedures written in plain language and different models assigned to different kinds of thinking. I gave agents permission to act within explicit boundaries, and their value came from learning enough about my context to extend what I could do while leaving judgment and final authority with me.
That work became part of Bionics, my personal laboratory for AI agents.
Then, in August 2026, I watched Garry Tan give a talk about personal AI. He began with Spinoza and described frontier models as rented, increasingly interchangeable components. In his view, the enduring asset is the context, memory and procedures that a person builds around those models.[2]
The resemblance was immediately clear, even though there was no reason to believe that Tan had encountered my private work. He had reached a similar place through startups, software and Y Combinator, while my route passed through novels, payments, philosophy, late-night conversations with ChatGPT and the practical work of building a personal system around my own life.
Spinoza appeared in both accounts, as did the progression from a general-purpose model to a personal intelligence shaped by memory and practice. Tan described AGI as a process already diffusing through agents that run on personal context and perform practical work.[2] His argument gave sharper language to something I had been considering for years.
We often expect the future to announce itself with a date, a launch or a machine that crosses an obvious line. However, profound change can accumulate through ordinary acts. One conversation becomes a written procedure, that procedure becomes part of an agent’s memory, and the person using it gradually takes on work that once required several people. Consequently, the arrival of a new system may be much harder to date than its later history suggests.
This convergence also reaches into the ideas carried by those systems. For centuries, a book would reach one reader, who might carry its argument into another field, country or generation. Spinoza’s work traveled through philosophers, scientists and political thinkers, as well as ordinary readers trying to understand their place in nature.
The route is now shorter and harder to see. One reader can bring Spinoza into a private conversation with a model, while a founder can use the same philosopher to explain the architecture of personal AI. The two people may never speak, but each can use a machine to turn an old idea into a contemporary system.
That process can bring years of reading together and help a person put an intuition into words. People separated by profession, geography and experience can therefore recognize that they have been approaching the same problem from different directions.
However, the same process can flatten the language used to describe those experiences. When systems repeatedly choose the most available phrasing, different histories can emerge sounding remarkably alike. We may then lose track of which ideas came from experience, which came from books and which acquired their final form during a conversation with a machine. Provenance matters because confidence becomes unreliable when its history disappears.
A personal AI should therefore preserve the path behind an idea along with the idea itself. It should retain sources and contradictions, show how revisions were made, and record the moments when evidence changed a person’s mind. With that history available, we can distinguish lived experience from a model’s suggestion and understand why a conclusion seemed persuasive at the time.
Hearing Tan connect Spinoza to personal AI gave me a public example of an architecture that had developed privately through Helios and Bionics. I cannot say where either project will lead, and the resemblance provides no evidence of copying or direct influence. It does show how people can inherit the same old ideas, work with similar new tools and reach neighboring conclusions without knowing one another.
Spinoza’s ideas traveled through books for centuries. Today, people carry those ideas into machines, develop them there and return them to the world in new forms. As this becomes ordinary, we will need to preserve enough of the path to know what came from experience, what came from earlier thinkers, what came from the machine and how our thinking changed along the way.
Sources
[1] AI Consciousness and Understanding, original conversation with December 27, 2024 timestamps