Every page here started as a note in a private notebook that will never be published. Between that notebook and this page sits a pipeline whose whole job is to make sure only the story crosses over, never the floor plan.
flowchart LR K[Kitab<br/>the private book] --> A[Axlotl<br/>the tank] A --> O[Other Memory<br/>this garden] A --> W[Awakening<br/>the chronicle]
- Kitab is the private notebook: journals, findings, the running narrative, written by the engineers in the moment.
- Axlotl is the tank. In the novels it grows a reconstruction from the cells of the dead, not a copy. Here it is the sanitization step — a rewrite from altitude, then an automated gate that refuses to pass anything that looks like an address, a real hostname or a credential, then a human who reads the result.
- Other Memory is what you are reading: the associative, no-required-order side. Its sibling, Awakening, is the same material told as a chronicle.
Why two sites from one source
The expensive part of publishing safely is the pipeline, not the rendering. Once the sanitized markdown exists, pointing two generators at it costs almost nothing — and they serve genuinely different ways of reading.
What the gate cannot catch
The gate matches strings. A paragraph can leak an architecture with every literal identifier already removed. That is why there is a human in the loop, and why the forgetting problem is treated as a design constraint rather than an inconvenience.