Every product screen re-cut per the Cutover ledger (2b–2f), as live HTML at real breakpoints. Product class is Porcelain Light by default; dark is a moment — the score reveal. All gold has left the product: eyebrows, verdicts and chips are teal or ink. Every KPI carries its data-trust state in words. The U6 re-skin implements exactly this on the real components.
Ledger 2d: named greeting · delete-account card leaves Home · trust states on every KPI · no gold anywhere · empty tiles get honest words, never fabricated numbers.
You've moved up within Cruising this week.
You're at step 3 of 5 · about 7 minutes left
Resume lessonMobile nav: five items — Progress folds into More (ledger P2 fix, harness selectors updated same change).
Ledger 2b + founder ruling 2026-08-18: the one dark surface in the product, and the verdict is a speedometer, not a number — four zones (Orienting · Handling · Cruising · Fluenting), needle position, no numerals. Zones grade like letter bands: defensible in an evolving field where an exact score isn't. The gold needle is the ruled gold-as-pointer role — the one gold allowed on a learner surface. Exact scores stay internal + Console. "Sample" chip whenever the verdict isn't the learner's own.
You're ahead of where you think you are — and stalled in a fixable place: delegation. You re-do work the machine already did.
Ledger 2e: level rows with teal state encoding (done = teal fill, current = teal tint), mono badges, skeleton loading instead of a bare sentence.
Ledger 2e: ONE takeaway rendering (the right rail owns it — chat body, mission card and "durable move" duplicates collapse) · human lesson numbers ("Lesson 4", never "Lesson 00") · no internal chips in learner view.
Ledger 2f: scenario card contrast fixed at the token level · no internal config as chat ("Mode: Pure Simulation…" gone) · rubric rows use trust-state hues with words · attempt labels human ("Keep iterating", never keep_iterating).
Good constraint. The model hedged on "skeptical" instead of citing. Tighten the instruction and resubmit.
Ledger 2f: next-best-action card in plain words (rule enums never render) · movement in zone language ("moved up within Cruising", never points) · readiness zone agrees with Home — same source, same trust state.
It's the next step in your path, and it feeds the delegation gap your assessment flagged.
Start Lesson 5Ledger 2f: Clerk appearance themed off tokens (the stray auth blues die) · anonymous pages get a real sign-in prompt, never a blank shell · marketing chrome, not app chrome, for anonymous visitors.
Welcome back. Pick up where you left off.
Continue with GoogleNo account yet? Start with the assessment — no signup needed.
One token file drives everything above — light product, dark score moment, trust states, teal action — per aina-tokens-all-themes.css Themes 1 + 5 and the colour-role rules. The U6 re-skin applies these classes to the real Astro/React components in route-sized batches on this same dev rig, founder sight per batch.