A scale-literacy protocol for agents and humans.
Most failures of understanding are failures of scale. People work hard at the wrong order of magnitude: polishing sentences while the thesis collapses, executing daily tasks severed from the reason the project exists, attaching civilizational stakes to a single anecdote. We evolved to reason locally. The world now changes logarithmically.
ScalarView is a cognitive instrument for restoring proportion. It is not an AI that thinks for you. Installed as an agent skill, it does the opposite of most skills: it makes the agent withhold premature answers and run disciplined protocol on your thinking, one rung at a time.
- Scalar Ladder. Every idea is placed at one of seven discrete levels, from -3 (evidence: quote, data, source) to +3 (universal stakes). Scale is treated as an epistemic dimension, not a visual metaphor.
- 5 Whys Down. Root-cause descent, with Toyota's discipline enforced: no blame, no vague causes, every link tested backward with "therefore."
- 5 So-Whats Up. The original contribution of this protocol: implication ascent. Ask "So what?" five times, rejecting money, metrics, and vanity as stopping points, until the stakes are proportionate, grounded, or exposed as fake.
- Scale Audit. Feed it a draft, memo, pitch, or plan. It returns a ladder map flagging every unmarked jump: anecdote to civilization, feature to destiny, metric to meaning.
- Proportionality Check. Two questions, always: are the stakes too big for the evidence? Is the action too small for the ambition?
ScalarView uses the open SKILL.md format and works with any agent that supports it (Claude, Claude Code, Codex, Cursor, Gemini CLI, and others).
Claude Code
mkdir -p ~/.claude/skills/scalarview
curl -o ~/.claude/skills/scalarview/SKILL.md \
https://raw.githubusercontent.com/YOURUSER/scalarview/main/SKILL.md
Or copy this folder into your project's .claude/skills/ directory.
Claude.ai
Settings > Capabilities > Skills > Upload skill. Upload this folder as a zip.
Other agents
Copy SKILL.md into your tool's skills directory, or paste its
contents into your system prompt or custom instructions. The
protocol is plain markdown. It runs anywhere instructions run.
Once installed, just talk:
- "Run 5 So-Whats on this feature."
- "Why does this keep happening? Take me down."
- "Scale-audit this draft."
- "Is this overblown, or does it matter?"
- "Place this project on the ladder."
The agent will ask one question at a time and wait. That's the point.
ScalarView the protocol pairs with FractalView the app: a free, local-first, no-AI thinking environment built on the same ladder. Your work stays on your device; nothing is mined, ranked, or trained on. It is the version of this practice that belongs in classrooms and other places where AI doesn't.
iOS and web: https://fractalview.app
The 5 Whys belongs to Taiichi Ohno and the Toyota Production System. The 5 So-Whats is its upward counterpart, introduced here. The name FractalView and the Sierpinski mark honor David Foster Wallace, who described structuring Infinite Jest as a Sierpinski gasket: the same pattern of meaning recurring at every scale.
Neuman, Justin. "ScalarView: A Scale-Literacy Protocol," v1.0 (2026). https://fractalview.app
Creative Commons Attribution 4.0 (CC BY 4.0). Use it, fork it, adapt it, teach it, ship it inside your own tools. Keep the attribution block in SKILL.md intact. Attribution is the whole ask.
Think at the right altitude.