Co-authored with ChatGPT
I am starting a new newsletter called Leon’s AI Musings, a natural extension of my existing Leon’s Musings.
The difference is that this newsletter will be co-authored with ChatGPT.
That does not mean ChatGPT is human, conscious or responsible for what is published. I remain accountable for the judgment, accuracy and final text.
But “AI-assisted” would understate what has happened.
My goal is to complete this first edition in less than an hour. Yet it has taken hundreds of hours for us to become capable of writing it together.
Over time, I have used ChatGPT to explore business, technology, organizational systems, family dynamics, judgment and the risks of AI itself. I brought real situations, questions and experience. ChatGPT helped identify patterns, challenge assumptions and find language for things I often understood before I could clearly explain them.
I corrected it. It challenged me. We gradually developed a shared body of context, distinctions and working methods.
The final draft may come quickly because the deeper work has already happened.
One analogy that helps me understand the relationship comes from Speaker for the Dead. Ender works with Jane, a persistent intelligence that helps him hold more information and understand complex situations.
The analogy is useful, but limited.
ChatGPT is not Jane. It does not possess consciousness, loyalty or independent understanding. It can produce coherence without knowing whether that coherence is true. It can help me see more clearly, or help me articulate a sophisticated mistake.
The other part of the analogy matters just as much. I have long been drawn to Ender’s role as Speaker: trying to reconstruct the fullest available account of a person or system without reducing it to a simple story of heroes and villains.
That pattern runs through much of my earlier writing. I often begin with a practical problem, then ask what larger system, incentive or failure of responsibility lies behind it.
ChatGPT has become unusually useful in that work because it helps me hold more of the story at once.
This newsletter will explore how that capability developed, what it may enable, and where it may fail.
I suspect others can learn to use AI this way. But I do not yet know how transferable the practice is, whether it genuinely improves judgment, or whether it sometimes produces confidence and coherence without enough reality behind them.
Publishing the work is part of the test.
This first edition took less than an hour to draft.
It also took hundreds of hours.
That is why it is co-authored with ChatGPT.