A dense, plain-English landing — rot-first hero, metrics strip, feature deep-dives with real product mockups, pay-once pricing, and an honest FAQ, on the TVA amber-CRT theme — converts better than the lo-fi prototype landing: more installs, Learning Kit intent, and calls.

Why this hypothesis

The current landing is lo-fi prototype branding: the copy leads with mechanisms ("evidence layer", "contracts") and the visual identity is a placeholder warm-paper theme. The bet has two halves that ship as one variant:

Copy — adopt the dense product-landing narrative arc (imperative rot-first hero, honest metrics strip, feature deep-dives with real product mockups, founder credibility, pay-once pricing anchor ~~$299~~ $249, build-vs-buy hours table, permission-granting FAQ) translated to plain English: Systemix stops your design system from rotting.
Theme — the TVA · Amber Mainframe design system (warm amber phosphor, soft CRT, light+dark). The terminal aesthetic is the product story: the deep-dive mockups render as the loop's own console.

Succeeds landing-ai-native-ds-2026-07 (same seam, same hero flag). This is ship-and-compare, not a split — at current traffic a real A/B cannot resolve, so the read is directional per-visitor rates before vs after.

Given

A pre-PMF founder (or anyone who owns a design system) lands on Systemix from a feed and gives the page five seconds. Today they must decode category jargon before knowing what Systemix does for them.

Measurement notes

Primary: install_command_copied per visitor (activation — the cleanest friction-free signal, per the SEO/GTM brief).
Secondary intent: kit_requested (revenue), audit_requested (pipeline), book_a_call (team). All existing events survive the rebuild verbatim, with the same location / variant / persona / section props.
The landing-hero flag keeps control / variant_b arms with the new copy pair ("Your design system rots. Systemix fixes that." vs "Your design system, but it learns.").

Decision criteria

promote: install copies per visitor rise meaningfully over the prior window, or the first kit_requested arrives → the rot-first framing becomes the permanent spine; iterate inside it.
iterate: engagement flat but time-on-page/section depth improves → the structure works, the words don't; tighten the hero pair, keep the arc.
kill: no lift anywhere → the dense product-arc doesn't fit this audience; revisit positioning (rot vs learning-loop lead) before touching layout again.

Changelog

2026-07-08 — Revision round from user testing (same bet, same metric): "the loop" banned from marketing copy (automation / AI-native design system that learns from signals / agentic-feed context-layer framing + the "one architecture, every audience" section); dense soft-grid layout (hairline frame, 3-col hero, stack strip, statement); CRT scoped to screen surfaces; dark body text warm white (spec v1.1); build-vs-buy rows expandable; Home dogfood made real (drift-report script + feed, runner worktree guard, full sidebar tree incl. design/ spine + skills/workflows catalog); app components swept onto the warm semantic tokens. Evidence read after this date reflects the revised page under the same bet.