Skip to content

docs: explain the token economics of surfaces#182

Merged
benvinegar merged 1 commit into
mainfrom
docs/token-economics
Jun 29, 2026
Merged

docs: explain the token economics of surfaces#182
benvinegar merged 1 commit into
mainfrom
docs/token-economics

Conversation

@benvinegar

Copy link
Copy Markdown
Member

What

The "Faster, and fewer tokens" bullet in the README asserted the benefit but never said why. This makes the case concrete:

  • Sharpens the bullet with the mechanism: a standalone HTML document re-sends its whole design system on every render, while a surface sends only the content and the viewer supplies the chrome — fewer tokens, and a faster draw (output tokens are written one at a time). It links down to the new section.
  • Adds a short ## Token economics section with a measured before/after table.

On the numbers

The percentages come from a generate-both-ways measurement: each artifact was authored as a full standalone HTML document and as the equivalent surface, then the output was tokenized and compared. They're rounded and stated modestly — the spread is the honest part (structured surfaces like diagrams/JSON/tables save ~80–90%; an interactive HTML mockup still saves ~40% by inheriting the viewer's theme rather than shipping its own).

Docs-only; empty changeset (no release).

🤖 Generated with Claude Code

The "Faster, and fewer tokens" bullet asserted the benefit without
saying why. Replace it with the concrete mechanism — a standalone HTML
document re-sends its whole design system on every render, while a
surface sends only the content and the viewer supplies the chrome — and
add a short "Token economics" section with a measured before/after
table (output tokens, generate-both-ways).

Docs-only; empty changeset (no release).

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
@benvinegar benvinegar merged commit bf91558 into main Jun 29, 2026
9 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant