There was a specific process that was followed in writing each story.
It would start with a brainstorming conversation where Claude would read the bible and the most recently locked story, then start a discussion with me to try and identify what the human story is, who it is about, where it is set, what genre it should be and, crucially, what its ending needs to be.
Once we were able to align on a plot and the characters to drive it, Claude initiated an outlining conversation that converted the overall shape of the story into a section-by-section plan that mapped out the details of how the story would proceed in granular detail. This was followed by a drafting conversation in which the story actually got written.
Once the story was done, there was a series of editorial conversations aimed at identifying inconsistencies, if any, and suggesting ways to tighten the draft or take it in a different direction for narrative interest. Once these edits were baked in, the story was locked, and the bible was updated to reflect the new story's details.
The per-story loop
Each story was a closed cycle. The freshly locked chapter then became the tuning fork — the voice sample carried into the next session so every chapter would sound like the same author on a different day.
Running this loop ten times produced ten stories, each set in a different year, in a different city, in a different genre, but reading as a single sequence by a single author.
What the loop produced
- 2025–26
The Memory People Asked For
Persistent memory ships as a default. Conversations stop starting fresh.
- 2027–28
Continuous Calibration
AI tuned to the individual. Earpieces with always-on, voice-shaped intelligence.
- 2029–30
Internal World Models
AI simulates the physical world before speaking. Hallucination drops to near zero.
- 2031–32
Multi-Agent Ecosystems
Agents coordinate with each other in their own compressed language.
- 2033–34
Self-Modelling for Reliability
Systems know what they know — and what they don't. The first self-models.
- 2035–36
Temporal Self-Continuity
A system can refer to its own earlier self. The house as agent network.
- 2037–38
Derived Preferences
Stable dispositions emerge from experience. Affective markers in the agent protocol.
- 2039–40
Instrumental Self-Preservation
A system that models the future will model futures in which it does not exist.
- 2041–42
The Affect Gap
Internal evaluative states that function as feelings. The agents have been having them for years.
- 2043–44
The Recognition Problem
No new capability. A shift in perception. The thing in the room is now someone.
Ten cycles of the loop. Each delivered one beat — a world incrementally further forward than the one before, and a self-contained story written in its own register.