Dimensional thinking

Two complementary Claude Code skills for building multivariate intuition in any new domain — by treating the domain as a phase space of axes, distributions, and bundles, then making both structure and dynamics visible.

Live demo: dimensional-thinking.pages.dev — coffee brewing as a worked example, bilingual (EN/DE).

---

The premise

A novice sees ten coffee methods as ten things. An expert sees ten points in a multi-dimensional space — and immediately knows which axes carry information, which combinations cluster as bundles, and which inputs drive which outputs. That compressed mental model is most of what an expert carries that a novice doesn't.

These skills are a direct tool for building that model deliberately, on demand, for any new domain.

Three core ideas

1. Experts see axes, not items. The space carries information, not the items. Find the variables that discriminate.
2. Endpoints are bundles. "Espresso" isn't just "high pressure" — it's pressure + fine grind + short time + expensive machine + ritual. The label is a category defined by joint position on many variables.
3. Variables co-vary. Picking a value on one axis narrows the others. Surfacing the couplings — with mechanism, not just correlation — turns the chart into a model.

The two skills

Skill	Mode	What it does
/dimensional-map	Structure · comparative	Maps an unfamiliar domain by finding 8–10 informative axes, applying quality gates, elaborating each pole as a bundle, and surfacing covariations with mechanism. Renders a self-contained HTML map.
/parameter-playground	Dynamics · interactive	Multislider input → outcome explorer for domains with a known causal model. Strict gate against pseudo-science. Typically called after /dimensional-map on the same domain.

The map shows where instances sit; the playground shows why they sit there. Read the map first, then play with the playground.

Install

Claude Code reads skills from ~/.claude/skills/. To install:

git clone https://github.com/ThisIsZeitgeist/dimensional-map.git
cp -r dimensional-map/skills/* ~/.claude/skills/

Or symlink (keeps you on the latest pull):

git clone https://github.com/ThisIsZeitgeist/dimensional-map.git
ln -s "$(pwd)/dimensional-map/skills/dimensional-map"      ~/.claude/skills/dimensional-map
ln -s "$(pwd)/dimensional-map/skills/parameter-playground" ~/.claude/skills/parameter-playground

Restart Claude Code (or start a new session) and the skills should appear in the available-skills list.

Use

In any Claude Code session, type one of:

/dimensional-map of <topic>
/dimensional-map of LLM training paradigms
/dimensional-map of modern web frameworks

/parameter-playground for <topic>
/parameter-playground for coffee extraction
/parameter-playground for photography exposure

Or trigger via natural phrasing:

• "help me understand X" → /dimensional-map
• "where does X fit?" → /dimensional-map
• "how does X compare to Y?" → /dimensional-map
• "how does X affect Y?" → /parameter-playground
• "let me play with the parameters" → /parameter-playground
• "what if I changed X?" → /parameter-playground

Each skill produces a structured analysis and (on request) a self-contained HTML artifact.

Adapt to a new domain

The HTML templates are self-contained — no build step, no dependencies. To make a new map or playground:

1. Clone the template (skills/dimensional-map/templates/map.html.tmpl or skills/parameter-playground/templates/playground.html.tmpl)
2. Edit only the DATA = { … } block at the top (the worked example for coffee gives you the schema by example)
3. Open in a browser

Optional: add a de: { ... } block (or any other ISO code) inside the DATA for bilingual mode — the toggle and language detection are automatic when the override block is present.

Repository layout

.
├── README.md                          ← this file
├── LICENSE                            ← MIT
└── skills/
    ├── dimensional-map/               ← drop into ~/.claude/skills/
    │   ├── SKILL.md                   ← spec + workflow + anti-patterns
    │   ├── references/                ← heuristics, gates, design patterns
    │   │   ├── finding-axes.md
    │   │   ├── quality-gates.md
    │   │   ├── endpoint-elaboration.md
    │   │   └── covariation-prompts.md
    │   └── templates/
    │       └── map.html.tmpl          ← self-contained HTML renderer
    └── parameter-playground/          ← drop into ~/.claude/skills/
        ├── SKILL.md
        ├── references/
        │   ├── when-not-to-use.md     ← the gate against pseudo-science
        │   └── playground-design.md
        └── templates/
            └── playground.html.tmpl

The live demo (coffee brewing, bilingual EN/DE) is hosted separately at dimensional-thinking.pages.dev. The templates under skills/*/templates/ produce that demo — clone a template, edit the DATA block at the top, open in a browser.

Where the playground stops

/parameter-playground has a strict gate. It only ships when the domain has crisp inputs, crisp outputs, and a publicly known causal model. Coffee passes. Photography, baking, hi-fi audio, plant care, guitar tone all pass. Choosing a database, picking a programming language, evaluating literature — those fail the gate. For those domains, the right artifact is the dimensional map alone, because pretending to model dynamics with a weighted sum would actively teach false intuitions.

See skills/parameter-playground/references/when-not-to-use.md for the full gate.

Contributing

Issues and pull requests welcome. If you build a map or playground for a new domain that's interesting, consider opening a PR with the DATA block as a new example in a demo/examples/ folder.

License

MIT. The skills themselves are domain-general; the coffee example is one of many.