Layout Type 01
Comparison
Side-by-side columns. Each column has equal visual weight — one preattentive badge differentiates the "best" option.
01A
3-Column Tool Comparison
Gestalt similarity: all columns same shape → parallel structure → easy scanning
Claude Code · Tool Comparison
CLI vs Browser vs Agent: Which Builds for You?
What each tool actually does when it's working
CLIClaude Code
Reads your codebase, plans before touching files, handles complex multi-file changes.
Git-aware commits and branches
Runs bash, installs packages
Best for: production codebases
Best for: existing projects
Browserv0 / Lovable
Visual-first generation. Fast from spec to deployed app, ideal for greenfield UI work.
Best for: new products fast
Best for: new apps, landing pages
AgentOpenClaw
Runs autonomously on a server, 24/7. Uses context and memory to make decisions without you.
Persistent memory across sessions
Telegram message & approval loop
Best for: automated pipelines
Best for: ongoing autonomous work
Decision rule: Start with Claude Code for existing code. Start with v0/Lovable for new products. Graduate to OpenClaw when the work runs better without you in the loop.
Layout Type 02
Tiers / Levels
Vertical progression. Size and indentation signal hierarchy. Number badges are the preattentive anchor — eye lands there first.
02A
4-Level Progression
Continuity: items aligned in column → eye reads as sequential progression
AI Automation · Framework
4 Levels of AI Automation
Pick the level that matches how much AI needs to own
1
Output
AI gives you answers
You ask, AI responds. You still route, decide, and act. Best for research, drafting, first-pass thinking.
ClaudeChatGPTPerplexity
2
Routing
AI output triggers other systems
AI runs in a fixed workflow. Invoice filed → CRM updated → task created. No judgment required.
n8nMakeZapier
3
Judgment
AI decides what to do next
Context changes what happens next. Lead priority judged, personalized follow-up drafted. Memory and reasoning required.
OpenClawCustom agents
4
Self-Learning
AI improves its own strategy
Reads outcomes, adjusts playbook. Noticed Bluesky converts 2× → shifted focus autonomously. Requires feedback loops + guardrails.
OpenClaw + memory
Layout Type 03
Process / Steps
Numbered horizontal or vertical flow. Arrows are connectors, not decorators — remove if the sequence is already clear from numbers.
03A
4-Step Horizontal Flow
Preattentive: numbered circles are the first visual anchor. Text is secondary.
OpenClaw · Setup Guide
4 Stages to a Production-Ready Agent
"Done" means all four stages working, not just the gateway
1
Gateway Running
Install Node 22+, run the onboard wizard, verify server starts on your port.
openclaw status → running
›
2
Identity Configured
Write SOUL.md (constraints), AGENTS.md (scope), USER.md (context). Without this, AI improvises at 3am.
3 files exist + filled
›
3
Telegram Paired
Pair via the approval code: openclaw pairing approve telegram [CODE]. Test with a direct message.
message → agent replies
›
4
First Cron Firing
Create workspace → add playbook → schedule job → verify UUID. Let it run unattended once.
job fires → output logged
Layout Type 04
Statistical / Hero Number
One dominant number anchors the entire graphic. Supporting stats are secondary. Data-ink principle: the number IS the design.
04A
Hero Stat + Supporting Metrics
Focal point: 72px number is a preattentive attribute — eye finds it in under 200ms
AI Coding Audit · Key Finding
8 Tools, Same Spec, Same Failure
What the 3-stage technical audit revealed
100%
of AI-built pages failed social SEO metadata
8 tools tested · Build to Launch audit · March 2026
3
tools lost mobile nav entirely
1
tool crashes on load (runtime error)
3min
Claude Code build time (fastest)
Stage 1 tells you who finished fast. Stage 3 tells you who's actually production-ready.
Layout Type 05
Bento Grid
4×2 grid of equal-weight cards. Every card answers the same question: what + why + what to do. Proximity groups cards — no inter-card borders needed.
05A
8-Card Reference Grid
Gestalt proximity: spacing within cards tighter than spacing between cards = grouping without borders
Claude Code · Plugin Review
8 Plugins. Which Ones Actually Earn Their Place?
Tested on real work — not just activated
Keep
brand-voice
Reads your brand docs, generates a 963-line guideline file, flags deviations in real time.
Always-on value
Keep
feature-dev
Plans architecture before writing. Stops you from building in the wrong layer from the start.
Prevents wrong arch
Keep
security-review
Caught path-traversal vulnerability (packId guard) in real code. Pays for itself once.
Caught real bug
Keep
marketing
Accurate competitor brief. Found "stuck-in-the-middle builder" positioning gap in your content.
Enable selectively
Skip
sales
Overlaps with marketing plugin. Mis-positioned your tier pricing in competitive brief.
Redundant
Skip
productivity
Redundant if you already have a task system. Adds overhead without replacing your workflow.
No added value
Cond.
commit-commands
Staging bugs on bracket-heavy paths. Test on your repo structure before committing to it.
Test first
Cond.
clinical-trials
Zero overhead when inactive. Extremely high value for healthcare/biotech — irreplaceable in domain.
Domain-specific
Layout Type 06
Feature Matrix
Rows = subjects, columns = criteria. High data-ink: checkmarks carry information, not borders. Color-coded check vs cross is preattentive.
06A
Tool × Feature Grid
Data-ink ratio: the ✓/✗ IS the content. Remove every pixel that doesn't carry meaning.
AI Tools · Technical Audit
Which Tools Actually Pass Stage 3?
Visual review hides what matters. These are the technical checks.
| Tool |
No console errors |
Mobile nav works |
OG tags present |
Optimal font loading |
Multi-file architecture |
| Claude Code | ✓ | ✓ | ✗ | ✓ | ✓ |
| Cursor | ✓ | ✓ | ✗ | ✓ | ✓ |
| v0 | ✓ | ✓ | ✗ | ✓ | ✓ |
| Lovable | ✓ | ✓ | ✗ | ✗ | ✓ |
| Firebase | ✓ | ✗ | ✗ | ✗ | ✗ |
| Gemini CLI | ✓ | ✗ | ✗ | SPA | ✗ |
| Bolt | ✗ | ✗ | ✗ | ✗ | ✗ |
| Replit | ✓ | ✓ | ✗ | ✗ | ✗ |
All 8 failed OG tags — social SEO is the universal blind spot. Bolt is the only crash-on-load failure.
Layout Type 07
Before / After
Two-panel transformation. Left = friction, right = resolution. The contrast IS the message — let it breathe, don't over-decorate.
07A
Transformation Comparison
Figure/ground: left panel muted, right panel slightly warmer — visual weight shifts to the "after"
Content Research · Workflow Shift
What a Research Day Looks Like Now
NotebookLM + Claude Code vs. the old way
✗Open 15 tabs, lose track of sources
✗Copy-paste quotes into a doc manually
✗Synthesize across sources by hand
✗Research session dies with the browser
✗Re-do same research two weeks later
✗Hallucinated stats from memory
✓49 sources indexed, queryable, cited
✓Ask questions, get grounded answers
✓5-layer synthesis in one query
✓Notebook persists across sessions
✓Query the same notebook next month
✓Every claim has a source citation
Layout Type 08
Timeline
Vertical spine with dot markers. The line is the visual backbone — all cards connect to it. Dates are secondary (enclosure in the card).
08A
Chronological Events
Continuity: the vertical line creates a path the eye follows. Cards are stops on the path.
AI Tools · Evolution
From ChatGPT to Autonomous Agents
The five moments that changed how we build
Nov 2022
ChatGPT changes the baseline
First time a general-purpose AI was genuinely useful for knowledge workers. The chatbox era begins.
Early 2024
Claude 3 + long context = real document work
100K token context means reading full codebases, legal docs, research corpora. Not just Q&A anymore.
Mid 2024
MCP opens the connections layer
AI can now talk to your tools — Gmail, Notion, GitHub, databases. The integration era begins.
Late 2024
OpenClaw / agentic platforms emerge
Always-on agents running on servers, with memory and Telegram control. Judgment replaces routing.
2025–2026
Claude Dispatch + self-improving systems
Remote GUI control + self-learning loops. Agents that update their own strategy based on outcomes.
Layout Type 09
2×2 Quadrant
Two axes create four zones. Dot position carries the data — labels and zones are secondary. The quadrant is the framework; content fills it.
09A
Positioning Map
Position is a preattentive attribute — viewer reads tool relationships before reading labels
AI Coding Tools · Positioning Map
Speed vs. Control: Where Each Tool Sits
Choose by what your project actually needs, not what looks impressive
← Less Control
More Control →
Faster ↑
↓ Slower
Fast + Less control
Fast + More control
Slow + Less control
Slow + More control
v0
Fastest, visual-first
🔥
Bolt · fast/crashes
Claude Code
Best production arch
Cursor
Best token arch
Lovable
Full interactions
Replit
Single-file output
Firebase
No mobile nav
Layout Type 10
Anatomy / Stack
Layered breakdown of a system. Each layer is numbered, labeled, and described. The stack metaphor encodes hierarchy — bottom = foundation.
10A
System Layer Breakdown
Enclosure + proximity: each row is a bounded object, stacked rows create the "whole system" gestalt
Claude Dispatch · Architecture
The 5-Layer Stack Behind Dispatch
The demos look effortless because the stack is invisible
1
Layer 1 · Direct API
MCP — Model Context Protocol
Bypasses the browser entirely. Direct API calls to Gmail, GitHub, Notion. Fastest path for structured data.
No screenshot loop needed
2
Layer 2 · DOM-native
Claude in Chrome
Reads and clicks the actual DOM. More reliable than screenshots for web-based tasks. No screenshot loop.
Best for web UIs without API access
3
Layer 3 · GUI fallback
Desktop Control (Computer Use)
Screenshot-based GUI interaction. ~60% reliable. Used when MCP and Chrome DOM both fail.
Last resort, not first choice
4
Layer 4 · Behavior
Skills & Plugins
Not required, but improve reliability. Brand voice, feature-dev, commit-commands add behavioral constraints.
Optional, but reduces stalls
5
Layer 5 · Environment
Claude Cowork + Claude Code
Session environment where tasks land. Cowork handles GUI tasks; Code handles file and coding tasks via terminal.
The execution surface
Layout Type 11
Roadmap / Phases
Horizontal phases with tasks inside each. The phase boundary IS the time boundary. Keep tasks short — each is an action, not a description.
11A
Phase-Based Learning Path
Continuity: columns read left-to-right as time → viewer scans phases naturally in sequence
AI Automation · Getting Started
Your First 30 Days with AI Automation
Three phases. Each builds on the last.
Set up Claude + Perplexity daily briefing
Run one research session via NotebookLM
Write one article with AI in the loop
Identify one recurring friction point
AI saves 1hr/day
Build one n8n workflow for a fixed task
Connect Gmail → Notion or a spreadsheet
Let it run for 5 days without touching it
One workflow runs itself
Install OpenClaw on Oracle ARM (free)
Write SOUL.md + AGENTS.md + USER.md
Create first scheduled cron job
Let it run overnight. Read the log.
Agent works while you sleep
Layout Type 12
Scorecard
Multi-criteria evaluation with dot ratings. Each row = one criterion. Verdict badge is the preattentive anchor — eye goes to green/red first.
12A
Plugin Evaluation Scorecard
Similarity: all rows same height and structure → fast scanning, no cognitive switching cost
Claude Code · Plugin Review
feature-dev Plugin: Scorecard
Evaluated on a real feature build — sentence exercises for kids-duolingo
Activation style
On-demand
Failure mode
Safe, not silent
Prevented wrong-layer architecture on the sentence-builder. One intervention = justified install.
Layout Type 13
Checklist
Grouped actionable items. Gestalt similarity: every checkmark same shape = fast scan. Chunked groups reduce cognitive load — reader sees 3 groups, not 9 items.
13A
Grouped Pre-Ship Checks
Enclosure: each group card = one cognitive unit — no extra separators needed
Claude Code · Pre-Launch
Before You Ship: The 9-Point Check
Run this before every deploy. Takes 8 minutes. Saves 8 hours.
Security
✓
Open your DB, inspect actual rows
Passwords hashed? PII encoded?
✓
.env not committed to git
Run: git log --all -- .env
✓
CSRF headers configured
Check response headers in DevTools
Testing
✓
Ran against prod-like DB
Not SQLite if prod uses Postgres
✓
Tested with real user inputs
Long strings, special chars, empty
✓
Mobile layout checked
375px, 390px, 430px widths
Shipping
✓
git commit before this session
If AI wipes it, you have a restore
✓
CLAUDE.md updated
Next session loads your rules
✓
Rate limits confirmed
API keys scoped to minimum access
Layout Type 14
Flowchart / Decision Tree
F-pattern reading: question chain runs down the left. YES branches exit right. The eye follows the path it needs and stops — irrelevant branches never compete for attention.
14A
Which AI Research Tool?
Selective attention: each branch resolves independently — reader processes only their path
AI Research Toolkit · Decision Guide
Which Tool Should Handle This?
Answer in under 30 seconds. Follow your path and stop.
What kind of task is this?
Code, files, or build something?
Claude Code
Edits files, runs bash, manages multi-step builds. Your default for anything in a repo.
Need grounded synthesis from specific sources?
NotebookLM
Upload your sources. Get citation-backed answers. Sessions persist — no rethinking needed.
Need real-time web data or broad search?
Perplexity
Live web search with citations. Best for market data, recent events, and fact verification.
Layout Type 15
Ranked List
Preattentive hierarchy: rank 1 is visually heavier before text is read. Bar width encodes magnitude at a glance — reader absorbs relative value without doing math.
15A
Top 7 by Real-World Impact
Similarity: bars are identical shape, only width varies — quantity reads in <200ms
Claude Code · Plugin Review
7 Plugins Ranked by Real-World Impact
Tested on actual builds — not demo prompts
01
feature-dev
Caught wrong arch
02
security-review
Real vuln found
03
brand-voice
Always-on writer
04
marketing
Brief in context
05
commit-commands
Test before commit
06
clinical-trials
Domain-specific
07
productivity
Redundant
Layout Type 16
Cost Bar / Horizontal Bar
Horizontal orientation lets long category labels breathe on the left. Single color family at varying intensities — quantity reads instantly, no legend needed.
16A
Token Cost by Task Type
Data-ink ratio: the bar fill IS the data — no axis labels, no gridlines needed
Claude Code · Token Economics
Token Cost by Task Type
Plan your sessions around what actually burns tokens
Quick question in context
One unfocused "fix everything" session costs as much as 8 targeted ones. Specificity is the cheapest token optimization.
Layout Type 17
Cycle / Circular Process
Gestalt closure: the dashed ring implies continuity without drawing arrows. Equal visual weight on all nodes communicates that no step dominates — the loop has no end.
17A
Content Engine Loop
Circular form encodes the concept — no explicit "repeat" label needed
Build to Launch · Workflow
The BTL Content Engine
Every article follows this loop — each step feeds the next
ALWAYS
ON
01
Research
NotebookLM + Perplexity
03
Write
Claude Code + brand-voice
04
Visualize
btl-infographic skill
05
Ship + Promote
Substack + OpenClaw
Layout Type 18
Funnel
Width IS the data — no axis needed. Color darkens toward the bottom to signal increasing rarity. Drop-off numbers between stages carry more information than the stages themselves.
18A
Builder Drop-off Funnel
Figure/ground: each stage is the figure; the shrinking width communicates loss without labels
Vibe Coding · Reality Check
Where Builders Drop Off
From idea to ongoing product — every stage loses people
Builders with an idea
100%
Everyone starts here
↓ 37% start building
Start building
37%
Most stop at setup
↓ first real bug hits
Survive first bug crisis
19%
AI loop burns them out
↓ ship to real users
Ship to real users
8%
Environment gap kills many
↓ keep shipping
The gap isn't talent — it's systems. CLAUDE.md, version control, and session discipline are what separate the 3% from everyone else.
Layout Type 19
Venn Diagram
The overlap zone is the thesis — everything else exists to make it legible. Muted fills let text in the intersection read clearly. Don't crowd the outer zones.
19A
Claude Code × NotebookLM
Figure/ground: overlap area is visually centered — eye lands there first, then reads outward
AI Toolkit · Tool Positioning
Claude Code + NotebookLM: What Only Works Together
Each tool has a lane. The middle is where the real leverage is.
Unique
File editing
Code execution
Git + bash
Terminal access
Long builds
Better together
Research → Build
Grounded briefs
No hallucination
Unique
Citation tracking
Audio overview
Multi-source merge
Persistent KB
Team sharing
Layout Type 20
Hub & Spoke
Centrality encodes hierarchy — hub visually dominates without text saying "this is the center." Spokes carry equal visual weight, communicating peer relationships rather than ranking.
20A
Claude Code Automation Stack
Gestalt proximity: spokes equidistant from hub = all tools are peers, not a hierarchy
Claude Code · Full Stack
The Claude Code Automation Stack
Every tool connects through the same center — the CLI session
Claude
Code
CLI
MCP Tools
Gmail, Notion, GitHub, Supabase — direct API access
Claude in Chrome
DOM-native browser automation — no screenshots
Vercel / CF
One-command deploys from terminal
Skills
Brand voice, infographics, commit rules
Memory
Persistent knowledge across every session
Claude Dispatch
Telegram-controlled, always-on agent server
Layout Type 21
Pyramid / Hierarchy
Width encodes rarity or magnitude — bottom wide = many/foundational, top narrow = few/advanced. Color gradient reinforces the hierarchy without additional decoration.
21A
5 Levels of Claude Code Mastery
Preattentive: width difference processed before text is read — hierarchy lands in <200ms
Claude Code · Mastery Framework
5 Levels of Claude Code Mastery
Most builders live at Level 2. The stack above it is what compounds.
5
Autonomous
Dispatch + memory. Self-running agents update their own strategy.
4
Workflow Automation
Skills + CLAUDE.md. Every session loads your rules — zero ramp-up.
3
Agentic Tasks
Multi-file projects. Bash, git, tests. Claude plans and executes sequences.
2
Codebase Editing
Reads files, makes targeted edits, understands architecture before touching code.
1
Prompted Answers
Ask questions, get answers. You decide, route, and act on outputs. Most builders start and stay here.
Data Type 22
Treemap
Area encodes magnitude — the biggest cell is the most expensive thing. No legend needed: darker = costlier. Where bento gives equal weight to all cards, treemap makes relative scale impossible to miss.
22A
Claude Code Token Budget by Task Type
Preattentive: area difference processed instantly — no numbers needed to see what costs the most
Claude Code · Token Economics
Where Your Token Budget Goes
Cell size = relative token cost. Darker = more expensive.
Full Codebase Audit
~40k tokens
Multi-file Refactor
~22k tokens
Feature Build
~18k tokens
Data Type 23
Bubble Chart
Three variables at once: x-position, y-position, and bubble size. Where a quadrant forces everything into 4 zones, a bubble chart shows continuous positioning across many items simultaneously — and size adds a third dimension of meaning.
23A
AI Builder Tools: Speed vs Control vs Cost
Preattentive: three variables (position × position × size) read before a single label — the trade-offs are visible in a glance
Claude Code · Tool Comparison
Speed vs Control vs Cost
Bubble size = relative token cost per session. Position = capability trade-off.
← Control
Speed →
Methodical
Fast
Low
High
Layout Type 24
Informational / Sectional
Linear sections with icons, headers, and body text — a visual explainer for complex concepts. Different from anatomy (which has a central diagram with callouts): this is sequential, text-heavy, and works for framework breakdowns and "what you need to know about X" summaries.
24A
What Claude Code Actually Does
Dual coding + enclosure: icon + text activates two recall paths; each section card is a bounded cognitive unit
Claude Code · Explainer
What Claude Code Actually Does
Five things that separate it from a chatbot in a text box.
📂
Reads your entire codebase
Not just the file you show it. Claude Code reads every file it needs to understand context before writing a single line. It knows what already exists.
File I/O
⚡
Runs real code in your terminal
Bash, Python, JavaScript — actual commands that execute, fail, and retry. Not suggested snippets you copy. It sees the output and adjusts.
Execution
🔌
Connects to your tools via MCP
Gmail, Notion, GitHub, Supabase. Claude Code doesn't just answer about your data — it operates on it. Read, write, send, deploy.
Tool Use
🧠
Remembers across sessions
Skills and memory persist. Every project rule, brand voice, and past decision loads automatically the next time you open the terminal.
Memory
🤖
Runs autonomously on a schedule
Claude Dispatch turns it into an always-on agent. It checks, fixes, and reports without you in the loop — triggered by time or by Telegram.
Automation
Sketch Type 25
Sketch / Annotation — Caveat
Full sketch register: Caveat for all text, dashed card borders, dot-grid canvas. Best for editorial/opinion content where the author's voice needs to be present in the design itself.
25A
When Claude Gets Stuck: 4 Moves
Caveat throughout — handwriting signals "author's voice, not data"
Claude Code · Troubleshooting
When Claude Gets Stuck: 4 Moves That Actually Work
In order. Each one resets a different thing.
1
Check context length first
If the conversation is long, Claude is working from a compressed summary of earlier decisions — not the actual code. Run /compact or start fresh.
→ /compact or new session
2
State the constraint explicitly
Claude can't guess what it can't see. If there's a rule, a pattern, or a reason — say it. "Don't touch X because Y" is better than hoping it remembers.
→ add to CLAUDE.md if it's recurring
3
Switch the output format
If it wrote bad code, ask for a plan. If it wrote a bad plan, ask for code. The format shift forces different reasoning. Often the second pass works.
→ "don't write code yet, just outline"
4
Show an example of what you want
One concrete example is worth more than three paragraphs of description. If you have a file that already looks right, share it. Claude will match the pattern.
→ "make it look like this file"
Sketch Type 26
Clean Handwriting — Kalam + Patrick Hand
Kalam for headlines (structured, clear), Patrick Hand for body (natural flow, very readable). Warmer than Inter without losing legibility. Works for how-it-works explainers.
26A
The 3 Layers of a Good Claude Code Project
Kalam headline weight vs Patrick Hand body weight — two registers, one voice
Claude Code · Project Setup
The 3 Layers of a Good Claude Code Project
Each layer does a different job. Skip one and the others have to compensate.
01
The Foundation
CLAUDE.md stores the rules Claude should never have to ask about: your brand voice, file conventions, what not to touch, and why. Write it once. It applies to every session.
Without this, Claude guesses. With it, Claude executes.
02
The Memory
Skills are reusable workflows — a skill for generating infographics, one for triage, one for commits. They load exactly when needed and don't clutter context otherwise.
Skills are the difference between repeating yourself and not repeating yourself.
03
The Execution
Subagents do the parallel work: one explores the codebase, one reviews, one writes. You orchestrate. The output is 3x faster than doing it sequentially yourself.
This layer only works well when the first two are solid.
Sketch Type 27
Friendly Notes — Indie Flower
Indie Flower: the most casual and approachable of the handwritten fonts. Reads like notes on a whiteboard. Best for checklists, signs, and low-stakes observational content.
27A
5 Signs Your AI Setup is Actually Working
Indie Flower — warm yellow-cream canvas, dashed separators, no hard structure
Build to Launch · AI Workflow
5 Signs Your AI Setup is Actually Working
Not metrics. Just things you notice.
1
You're attempting things you wouldn't have tried before
The work isn't faster yet — it's different. You're reaching for problems that used to feel too big.
scope shift
2
Your first draft is already 80% there
Not perfect, but close enough to edit rather than rewrite. The gap between output and done is small.
output quality
3
You've stopped re-explaining the same things
Brand voice, file patterns, don't-touch rules — they're in CLAUDE.md now. You wrote them once.
context is persistent
4
The correction loop is getting shorter
Fewer back-and-forths per task. When Claude misses, you know why — and it's usually something you can fix once.
feedback loop tightening
5
You're shipping before you finish describing
You started building before you had a full spec. The output helped you figure out what the spec should have been.
velocity is real
Narrative Type 28
Problem / Solution / Action
The only rhetorical format — designed to argue and persuade, not just inform. Three panels carry the reader through an emotional arc: conflict → evidence → action. Most common BTL content pattern.
28A
Why Most AI Setups Fail in 90 Days
Figure/ground — problem panel muted (gray border), evidence panel lavender, action panel full purple. Emotional arc encoded in color progression.
Build to Launch · AI Workflow
Why Most AI Setups Fail in 90 Days
And what to do instead
The Problem
You automated before you understood
Most builders connect an API, see a result, and call it a workflow. But automation amplifies your existing process — good or bad. If the manual version was messy, the automated version is faster chaos.
73% of AI tools adopted in Q1 are abandoned before Q4
The Pattern
What actually works
✗Automate first, understand later — always breaks
✓Document every step before touching the API
✓Only automate tasks you've done manually 10+ times
✓Start with the exception list, not the happy path
Your Move
The 3-question test
Before automating any task, ask: "Can I describe every step without pausing to think?"
If yes → automate it
If no → do it 5 more times first
Gallery Complete
28 Types Rendered
Comparison · Tiers · Process · Statistical · Bento · Matrix · Before/After · Timeline · Quadrant · Anatomy · Roadmap · Scorecard · Checklist · Flowchart · Ranked List · Cost Bar · Cycle · Funnel · Venn · Hub & Spoke · Pyramid · Treemap · Bubble Chart · Informational · Sketch/Caveat · Kalam+Patrick Hand · Indie Flower · Problem/Solution/Action
Each built from first principles: dual coding, Gestalt proximity, preattentive hierarchy, data-ink ratio, Z-pattern reading path.