Skip to content

marketing: make Marginal obvious in five seconds - #29

Merged
SignalLayerLabs merged 1 commit into
mainfrom
marketing/refactor-first-5-seconds
Aug 18, 2026
Merged

SignalLayerLabs merged 1 commit into
mainfrom
marketing/refactor-first-5-seconds

Conversation

@SignalLayerLabs

Copy link
Copy Markdown
Owner

Refactors MARGINAL's public surface around one immediate product message:

AI agents repeat work that changed nothing. MARGINAL catches it.

Marketing

  • new first-5-seconds positioning
  • visual no-progress loop demonstration
  • prominent GitHub star/install CTAs
  • clearer Earned Enforcement story
  • benchmark moved from headline to research/proof
  • redesigned README hero
  • dedicated shareable demo page
  • new social/Open Graph artwork

SEO

  • rewritten title and meta description
  • canonical metadata
  • Open Graph + Twitter cards
  • SoftwareApplication structured data
  • FAQPage structured data
  • search-intent FAQ content
  • sitemap and robots coverage
  • semantic heading hierarchy

Claim discipline

The existing exploratory benchmark remains published and unchanged in meaning:

  • 0/3 → 0/3 resolved
  • 24.93% observed token difference
  • no applied deny
  • pass_through
  • no causal savings claim

@SignalLayerLabs
SignalLayerLabs merged commit 1d84bd2 into main Aug 18, 2026
6 checks passed
SignalLayerLabs added a commit that referenced this pull request Aug 18, 2026
Aligns the canonical MARGINAL Killer Demo with the new public positioning already shipped in #29.

AI agents repeat work that changed nothing. MARGINAL catches it.

Changes
redesigns /demo/ around the no-progress loop
adds Same action / Same state / No new evidence visualization
separates Shadow Mode from Earned Enforcement
keeps deterministic allocation results clearly separated from runtime-governance illustration
preserves reproducibility and existing result semantics
regenerates the Codex runtime and provenance
Claim discipline
The deterministic demo remains explicitly:

not provider telemetry
not a production benchmark
not evidence that every agent workload saves compute
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