Every page, in one list
Three sections, 86 pages. Topics set out the approach and its
failure modes; the guides under each one work through a single specific problem end to end —
what causes it, the code that fixes it, and the check that proves it stayed fixed.
How the three levels of this reference relate
Three levels are drawn top to bottom. At the top sit the three sections, each covering one major surface of a synthetic-spatial pipeline: the architecture and governance around a release, the generation of spatial distributions, and the simulation of movement. Beneath each section sit its topics, which set out an approach in full — the concept, an implementation, how to validate it, how it scales, and how it fails. Beneath each topic sit its guides, each of which takes a single specific problem and works it end to end: what causes it, the code that fixes it, the check that proves it stayed fixed, and the edge cases that make the obvious fix insufficient. Alongside each level is the question it answers — what area is this, how do I do this, and why is this specific thing broken. The note underneath suggests entering at the level that matches what is already known: a guide when there is a symptom, a topic when there is not.
Section, topic, guide — three levels, three different questions
sections
one surface of the pipeline
3
"what area is this?"
topics
an approach, end to end
25
"how do I do this?"
guides
one specific problem
58
"why is this broken?"
where to start
a symptom in hand → a guide
a stage to build → a topic
a whole area new → a section
Every guide names the topic
above it in its first line, and
every topic lists its guides, so
any entry point reaches the rest.
Enter at the level that matches what is already known. The guides assume the topic's material and do not repeat it; the topics assume nothing
beyond the section.
Section, topic, guide: what each level answers, and where to enter depending on what you already know.
Topics and guides in each section
Three sections run down the vertical axis. For each, two horizontal bars show the number of topics and the number of guides beneath them. Architecture and governance carries nine topics and eighteen guides, covering contracts, seeding, privacy frameworks, differential privacy, adversarial testing, lineage, realism metrics, continuous integration and production readiness. Spatial distributions carries nine topics and twenty-two guides, covering point processes, tessellation, density surfaces, generative models, attribute correlation, network topology, and the memory and parallelism concerns of large grids. Trajectory simulation carries seven topics and eighteen guides, covering physics-based paths, agent-based mobility, Markov routing, noise and drift, map matching, stop detection and temporal synchronisation. The note underneath explains the ratio: roughly two to three guides per topic, because a topic that generates fewer specific, reproducible problems than that is usually a topic that has not been used in production.
Where the depth sits
0
5
10
15
20
25
pages
9 topics
18 guides
Architecture & governance
9 topics
22 guides
Spatial distributions
7 topics
18 guides
Trajectory simulation
Roughly two to three guides per topic. A topic that generates fewer specific, reproducible problems than that is usually one that has not been
run in production long enough to know what breaks — and the guides are written from what breaks.
Topics and guides in each section — roughly two to three guides per topic, written from what actually breaks.
What a topic page and a guide page each contain
Two outlines side by side. A topic page opens by framing the problem the approach exists to solve, states its prerequisites, explains the core concept, works through a step-by-step implementation, covers validation, performance at scale, and the ways it fails in production, and closes with common questions and links onward. A guide page is narrower and deeper: it names the root cause of one specific failure, gives the fix as working code, gives the verification step that proves the fix held, covers the edge cases where the obvious fix is insufficient, and links back to its topic. Both carry measured figures rather than illustrative ones — every chart on the site is computed from a real algorithm with a fixed seed, so the numbers in the captions can be reproduced. The note underneath states that consequence directly.
What each kind of page contains
a topic page
Problem Framing — what this exists to solve
Prerequisites — what has to be true first
Core Concept — the mechanism itself
Step-by-Step Implementation — working code
Validation & Testing — how you know
Performance & Scale — where it stops working
Failure Modes — how it breaks in production
FAQ and Related — questions and links onward
a guide page
Root Cause — why the failure happens
Fix — the change, as code you can run
Verification Step — the check that proves it
Edge Cases & Gotchas — where the fix falls short
Related — the topic above and its siblings
Narrower and deeper than a topic: one
specific symptom, worked end to end.
Every chart on the site is computed rather than drawn — a real algorithm, a fixed seed, and the parameters stated in the caption — so any
number quoted in a figure can be reproduced from the description beside it.
What a topic page and a guide page each contain, so you know what you are getting before you click.
Pipeline blueprints, scoping contracts, privacy-preserving generation, realism metrics, and CI/CD gating for synthetic spatial data.
Point processes, tessellation, density mapping, async grid execution — turning statistical constraints into coordinate-accurate geometry.
Markov routing, physics-based path generation, noise injection, multi-agent temporal sync — production mobility simulation patterns.
