No ceiling on your docs.

Markdown is a ceiling. Your thinking deserves more.

You write the doc. The numbers go into a static table because markdown can't chart them. The architecture goes into ASCII art because markdown can't render diagrams. The decision tree gets flattened into bullets. The CSV gets pasted as a code block. The doc ships. It gets archived. You never reopen it — because there's nothing to reopen into.

• Perf reports with raw numbers in a static table when they deserved a chart bound to the source data — edit the CSV, watch the chart move.
• Architecture docs that describe systems in prose because markdown can't show a clickable diagram with click-to-explain panels.
• Decision walks flattened into nested bullets because there's no way to actually walk a decision tree in markdown.
• Learning notes that read like transcripts. No glossary popovers, no inline quiz, no interactive concept map. Just paragraphs.
• Vision docs that get shared once and never reopened. The format sets the ceiling for how alive the doc can feel — and markdown's ceiling is flat.
• The alternatives fail differently. Notion and Observable live on someone else's server. Idyll requires authoring a separate markup language. Claude Artifacts disappear with the chat. None of them give you a self-contained, portable, archive-worthy artifact that runs from your filesystem.

Below is the same Q1 funnel rendered two ways — flip the tab and see what the format actually costs you.

Let the agent build the doc your thinking deserves.

inkling is a Claude Code skill. You describe the document in natural language — no markup, no slash commands. The agent designs the chassis for this specific doc, invents the structural shapes it needs, writes the styles, wires the interactions, parses your data, and renders a folder of HTML + CSS + JS + data you can open offline, archive forever, or share.

Every render must hit five axes. None are negotiable. If any axis falls below threshold, the agent fixes it before declaring done.

The rest of this page shows how it actually works — the primary flow, a typical session, what a doc folder looks like inside, and the prompts to paste into Claude Code to start your own. This page was built by inkling. Scroll to the bottom to see its own event log.

What happens between your message and the live doc

Click any step to see what the agent does, and what you see.

Six turns that build a real doc

A simulated conversation with the agent — friend talks on the left, the agent reports on the right with the files it actually touched. Notice you never see a slash command.

Anatomy of a doc folder

A doc isn't a file — it's a folder. Click any node in the tree to learn what each file does and why it's there.

Three docs inkling has rendered.

Each link below opens a real inkling artifact. Read the doc-source.json inside each folder to see the event log that produced it — provenance is part of every doc.

• Vision doc A project's vision rendered as an inkling artifact. Uses the starter visual vocabulary heavily; kept here as an honest reference for what the warm-notebook tokens look like and where templater drift can creep in (this render predates the originality contract).
• The inkling skill explained A learning-shaped doc with inline charts, glossary popovers, and click-to-explain wiring. Demonstrates behavioral invention at the runtime layer — parseCsv, renderBarChart, clickToExplain, glossary all written fresh.
• Full gallery → All three reference renders side-by-side with originality scores and one-line notes on what each one demonstrates.

Start your first doc in two messages

Open Claude Code in any folder. Paste the first prompt. After it replies, paste the second. That's it.