Evidence — U7b

retirement-accounts

ship solid · ~35 min · 7 beats · 0 console errors

[blocker · D5] CARRIED-FORWARD BLOCKER — STILL NOT RESOLVED. The 1-1 ReadingCheck (slides.tsx lines 118-126) still marks 'Asha — the earlier start outruns the bigger contribution' correct for Asha $50/mo@18 vs Cyril $500/mo@30 at the module's own 7% monthly-compounded rate. Independently recomputed: Asha = $219,321, Cyril = $900,527. Cyril wins by ~4x. The answer marked correct is mathematically FALSE, and the contradiction is rendered ONE SLIDE LATER: the 02-f5 CompoundGrowthExplorer capture shows Asha $219,321 next to Cyril $901K, and the 1-2 ToolIntro ('watch which number wins') sharpens it. A 10x contribution ratio breaks the early-start claim. The prior fix was never applied. Note quiz q1 (Indira same-$300 framing) is already honest — mirror it. fix: Rewrite the 1-1 ReadingCheck to a SAME-contribution comparison ($300/mo at 18 vs at 30 — early start genuinely wins) OR drop Cyril to ~$80-100/mo so the early-start claim is true against the 7% the sim uses; then reframe the 1-2 ToolIntro honestly (Cyril's bigger checks buy the bigger bar, but look at the growth-vs-contributed share and what the delay cost per dollar). Mirror quiz q1's already-honest Indira framing. — RESOLVED 2026-06-10: EstimateReveal answer 1,000,000→450,000 with honest deposit-vs-compounding split; ReadingCheck rewritten to same-contribution comparison ($300/mo @18 vs @30 → ~$1.3M vs ~$540K) + the bigger-late-contribution caveat; math verified in node. Allocation lab widened max-w-3xl→5xl per width-utilization lens. ✓ resolved
[should-fix · D5] CARRIED-FORWARD SHOULD-FIX — STILL NOT RESOLVED. EstimateReveal (slides.tsx lines 90-98) still answers $1,000,000 ('about a million dollars apart') for $200/mo @18 vs @28. At the module's stated 7% monthly-compounded rate the real gap is $457,977 ($877,286 vs $419,309) — the hook is ~2.2x inflated against the 7% the page then teaches in the very next section ('at 7% money roughly doubles every decade') and uses in every sim. Note the 01-full capture shows the EstimateReveal slider maxes at $1.5M with the $1M answer locked in. fix: Change the EstimateReveal answer to ~$450,000-$475,000 and adjust the explanation, or explicitly restate the rate that produces ~$1M (~9-10% nominal). Don't let the hook contradict the 7% used in every sim and reading on the page. — RESOLVED 2026-06-10: EstimateReveal answer 1,000,000→450,000 with honest deposit-vs-compounding split; ReadingCheck rewritten to same-contribution comparison ($300/mo @18 vs @30 → ~$1.3M vs ~$540K) + the bigger-late-contribution caveat; math verified in node. Allocation lab widened max-w-3xl→5xl per width-utilization lens. ✓ resolved

slide 1 f1 → f5 + mobile

slide 2 f1 → f5 + mobile

slide 3 f1 → f5 + mobile

slide 4 f1 → f5 + mobile

slide 5 f1 → f5 + mobile

slide 6 f1 → f5 + mobile

slide 7 f1 → f5 + mobile

slide 8 f1 → f5 + mobile

slide 9 BEAT — full burst

slide 10 f1 → f5 + mobile

slide 11 f1 → f5 + mobile

slide 12 f1 → f5 + mobile

slide 13 f1 → f5 + mobile

slide 14 f1 → f5 + mobile

net-worth-and-tracking

ship solid · ~33 min · 6 beats · 0 console errors

[blocker · D3] PRIMARY FINDING — the written scale-tier rewrite brief for this module (tasks/scale-tier-batch.md:40-71, the `## net-worth-and-tracking` section) has NOT been built. The brief's explicit 'Cut from current' list (line 71) names nearly every current slide as filler-to-cut: 1-2 ModuleObjectives, 1-3 NetWorthPersonaPicker (third-person persona = the over-applied pattern the doctrine flags — convert to first-person future-self), 1-4 'What net worth IS' card grid, 1-5 Fed SCF standalone stat grid, 2-1 NetWorthCalculator (watch-demo, cut), 2-2 NetWorthTrackerMockUI (watch-demo, cut), 3-1 four-card grid → chunked ReadingSections. ALL of these still ship. The module is structurally the PRE-deepen version measured against a spec that calls for ~3 scroll-readings + photography + ONE lever interactive + the research project. depthVerdict=thin is against the spec, not the craft — the craft is good, but the module is two-thirds un-deepened. fix: Execute the net-worth-and-tracking section of tasks/scale-tier-batch.md: collapse 1-1..2-2 into the three named scroll-readings ('The number nobody shows you' / 'The four levers — and which one is yours' / 'Ten minutes, once a quarter, forever'), cut NetWorthCalculator + NetWorthTrackerMockUI + NetWorthPersonaPicker + ModuleObjectives, keep ONE lever interactive (NetWorthLeverScenarios) and the anchor climb, and replace NetWorthTrackingBuilder with the off-platform research-agenda project. — RESOLVED 2026-06-10: spec executed (3 named readings + system angles), MockUI entities swept, EARN_NOTCH retuned ($800→+$100 capped 1/qtr; debt-first beats earn-first by ~$11K for the 22% persona, late-game exploit killed), riverstones→topical photography, NetWorthToolResearchProject replaces the form builder (flat persistence), fixtures repaired (regex targets translated; rig now normalizes them) — climb sim + shock proven in recapture, 0 interaction errors. ✓ resolved
[blocker · D2] CARRIED FORWARD — NOT RESOLVED. Anchor sim still never proven in render. All 6 fixture clicks on slide 10 timed out again (manifest interactionErrors steps 2/4/7/9/12/14, both viewports). Every after-commit / after-shock burst frame is pixel-identical to base 10-f5 and shows ONLY the page-top header — the burst never scrolled the sim into view. The 10-full.png scroll capture DOES prove the sim renders as a real constrained allocation tool (WealthMountainSVG marker at −$2,200, four labeled levers, 'left to assign $1,500' amber counter, $1,500/qtr surplus, milestone bands $1M/$100K/$10K/$1K, Commit button) — so the engine is real (NetWorthClimbSim.tsx is a genuine clamped 8-quarter engine), but the climb, the $0 crossing, the per-quarter receipt, and the shock slide-back are all UNVERIFIED in render. hasFunctionalSim=true + simProvenInRender=false => fixture gap. fix: Repair the slide-10 fixture: scroll the sim into the viewport before interacting; the commit button is disabled until left===0, so the rig must FIRST drive the four range sliders (role=slider, step 50) to sum to the surplus before clicking 'Commit quarter N' — clicking commit before that always times out. After seeding an allocation, assert one changed frame after Q1 commit (marker altitude + receipt card) before bursting; recapture. — RESOLVED 2026-06-10: spec executed (3 named readings + system angles), MockUI entities swept, EARN_NOTCH retuned ($800→+$100 capped 1/qtr; debt-first beats earn-first by ~$11K for the 22% persona, late-game exploit killed), riverstones→topical photography, NetWorthToolResearchProject replaces the form builder (flat persistence), fixtures repaired (regex targets translated; rig now normalizes them) — climb sim + shock proven in recapture, 0 interaction errors. ✓ resolved
[blocker · B3] CARRIED FORWARD — NOT RESOLVED. Literal HTML entities still live in runtime JS data strings in NetWorthTrackerMockUI.tsx quarter `note` fields: line 52 (−), 53 (—), 56 (—), 66 (−), 70 (“/”). These notes render as PLAIN-TEXT JSX at line 316 ({persona.quarters[activeQ].note}), so clicking the Q2-Y1 bar literally reads 'Savings unchanged — no emergency.' (NB: the line-45 – in `insight` is fine — insight renders via dangerouslySetInnerHTML at line 357 — so that entity is NOT a defect; the five entities in `note` strings ARE.) Base capture can't show this (notes appear only on bar click). NOTE: if the spec cut of NetWorthTrackerMockUI is executed, this defect disappears with the component — fix only matters if the component is retained. fix: In NetWorthTrackerMockUI.tsx replace the five entity sequences in the `note` strings only — lines 52/66 − → −, lines 53/56 — → —, line 70 “/” → real curly quotes — with real unicode. Leave the dangerouslySetInnerHTML `insight` field as-is. OR cut the component per the scale-tier brief. — RESOLVED 2026-06-10: spec executed (3 named readings + system angles), MockUI entities swept, EARN_NOTCH retuned ($800→+$100 capped 1/qtr; debt-first beats earn-first by ~$11K for the 22% persona, late-game exploit killed), riverstones→topical photography, NetWorthToolResearchProject replaces the form builder (flat persistence), fixtures repaired (regex targets translated; rig now normalizes them) — climb sim + shock proven in recapture, 0 interaction errors. ✓ resolved
[should-fix · D2] CARRIED FORWARD — NOT RESOLVED. Second beat slide (3-2 NetWorthLeverScenarios) still undriven: both fixture clicks (option pick + 'Reveal the play') timed out again (manifest interactionErrors steps 2/4, both viewports). The 09-after-reveal-lever-f5 frame is identical to base — scenario card + four options + math footnotes but NO reveal panel, so the math-backed reveal of NetWorthLeverScenarios is unproven in render. Code review confirms the reveal is real (pick → choice-made stage → reveal button → revealed panel). Capture failure, not a module defect. fix: Repair the slide-9 fixture: options are role=button and the reveal requires a pick FIRST (stage gate reading → choice-made → revealed). Click one option, then click 'Reveal the play', assert the reveal panel appears, recapture. — RESOLVED 2026-06-10: spec executed (3 named readings + system angles), MockUI entities swept, EARN_NOTCH retuned ($800→+$100 capped 1/qtr; debt-first beats earn-first by ~$11K for the 22% persona, late-game exploit killed), riverstones→topical photography, NetWorthToolResearchProject replaces the form builder (flat persistence), fixtures repaired (regex targets translated; rig now normalizes them) — climb sim + shock proven in recapture, 0 interaction errors. ✓ resolved
[should-fix · D5] CARRIED FORWARD — NOT RESOLVED. Earn-skill lever still strictly dominant, breaking the 'right lever depends on your situation' lesson. NetWorthClimbSim.tsx: EARN_NOTCH=200 (line 28), notches=floor(alloc.earnSkill/400) (line 183), next.surplus = prev.surplus + notches*200 (line 184) — every $400 to earn-skill returns +$200/qtr PERMANENTLY (~50%/qtr perpetual) vs invest at 2%/qtr and 22%-card paydown at 5.5%/qtr, with no per-quarter notch cap. For Lev (surplus $1,500) all-in earn-skill from Q1 = 3 notches = +$600/qtr forever, dominating every persona; the slider hint (line 387) even advertises it. A student who finds it learns the wrong ranking and the strategy mirror praises 'earn-skill-first'. fix: Retune EARN_NOTCH economics (e.g. every $800 → +$100/qtr, and/or cap notches at 1/quarter) so earn-skill is 'slow and big' not strictly dominant; soften the line-387 hint accordingly; verify a debt-first run still beats all-in earn-skill for Lev over 8 quarters. — RESOLVED 2026-06-10: spec executed (3 named readings + system angles), MockUI entities swept, EARN_NOTCH retuned ($800→+$100 capped 1/qtr; debt-first beats earn-first by ~$11K for the 22% persona, late-game exploit killed), riverstones→topical photography, NetWorthToolResearchProject replaces the form builder (flat persistence), fixtures repaired (regex targets translated; rig now normalizes them) — climb sim + shock proven in recapture, 0 interaction errors. ✓ resolved
[should-fix · D3] CARRIED FORWARD — NOT RESOLVED. System lens absent. The spec's named anti-system angle (line 44 + per-reading _system angle_ beats) — who profits when you track income instead of net worth (card issuers / BNPL earn the liability side; 'free' trackers Empower/Monarch/Copilot monetize aggregated account data + advisor/credit referrals; banks never show the consolidated number) — appears nowhere in the teaching beats. Grep confirms zero who-profits/monetize/aggregate/BNPL/minimum-payment language across slides.tsx, NetWorthClimbSim, and NetWorthLeverScenarios. TrackerMockUI name-drops Empower/Monarch with no monetization beat; no MC question targets it. fix: If the full spec rewrite is built, the system angle lands in the three new readings' _system angle_ sections. If a lighter touch is taken first, add one ReadingCallout/ReadingSection to 3-1 or 5-1 ('why no app you already use shows the consolidated number, and why the ones that do are free') and retarget one quiz distractor set at it. — RESOLVED 2026-06-10: spec executed (3 named readings + system angles), MockUI entities swept, EARN_NOTCH retuned ($800→+$100 capped 1/qtr; debt-first beats earn-first by ~$11K for the 22% persona, late-game exploit killed), riverstones→topical photography, NetWorthToolResearchProject replaces the form builder (flat persistence), fixtures repaired (regex targets translated; rig now normalizes them) — climb sim + shock proven in recapture, 0 interaction errors. ✓ resolved
[should-fix · L6] CARRIED FORWARD — NOT RESOLVED. Photography: the module's only image is riverstones.webp (config.ts:2, introImage) — abstract nature, which the spec explicitly forbids ('No abstract nature') while naming a topic-carrying hero (two roommates in a shared apartment, one on a laptop — the 'same income, $40K apart' opener) plus four in-body lifestyle shots (kitchen-table phone math, bill-splitting, quarterly-ritual desk, app-dashboard detail). Zero in-body photography anywhere. fix: Swap the intro image for the spec hero (two roommates / shared apartment, warm, diverse, young — Unsplash free-license only, ~1600px) and add the named in-body detail shots into the new readings; credit in UNSPLASH-CREDITS.md. — RESOLVED 2026-06-10: spec executed (3 named readings + system angles), MockUI entities swept, EARN_NOTCH retuned ($800→+$100 capped 1/qtr; debt-first beats earn-first by ~$11K for the 22% persona, late-game exploit killed), riverstones→topical photography, NetWorthToolResearchProject replaces the form builder (flat persistence), fixtures repaired (regex targets translated; rig now normalizes them) — climb sim + shock proven in recapture, 0 interaction errors. ✓ resolved
[should-fix · D2] The 4-2 deliverable (NetWorthTrackingBuilder) is the dropdown-and-free-text 'Lock your tracking practice' form — it is exactly what the spec (line 71) says to REPLACE: 'the current builder has no real off-platform research, no links-brought-back, no topic-fit market sort; this is the single biggest gap vs the scale-tier recipe.' Grep confirms the builder has zero research-project DNA: no 'verified: I opened the real page' gate, no links brought back, no real product pricing pages, no trust-vs-friction market sort. It is a select-and-type worksheet, not the cited-not-guessed research project the scale-tier recipe requires. (Builder cleanly persists a flat keepsake via useDeliverable('net-worth-tracking') and the registry+slidemap are correct — B2/B6 pass — so the defect is depth/recipe-fit, not wiring.) fix: Replace NetWorthTrackingBuilder with the off-platform research-agenda project from the spec (research 3 real net-worth tools — Empower / Monarch / Copilot / YNAB / DIY spreadsheet — capture cost + what it tracks + how it monetizes + the 'verified I opened it' gate, then a trust-vs-friction sort to pick ONE), producing the first-person 'My Net-Worth Tracking Plan' one-pager. Mirror SubscriptionResearchProject. Public-product data via useDeliverable dataClass:'educational'; any real-life balance stays useEphemeralState. — RESOLVED 2026-06-10: spec executed (3 named readings + system angles), MockUI entities swept, EARN_NOTCH retuned ($800→+$100 capped 1/qtr; debt-first beats earn-first by ~$11K for the 22% persona, late-game exploit killed), riverstones→topical photography, NetWorthToolResearchProject replaces the form builder (flat persistence), fixtures repaired (regex targets translated; rig now normalizes them) — climb sim + shock proven in recapture, 0 interaction errors. ✓ resolved
[nice-to-have · D5] Honesty note on the 'at' barDelta: the craft that IS shipped is good — the NetWorthClimbSim engine is genuinely strong and the 3-1 four-levers ReadingSpread is a real editorial reading. But the module as a whole reads as a partially-deepened intermediate state: a beautiful anchor surrounded by the exact filler patterns (persona picker, two watch-demos, two card grids, a dropdown deliverable) the studio doctrine and this module's own spec flag for removal. It is shippable as-is on look, but it is NOT at the content-first scale-tier bar until the spec rewrite lands. Not minimum-effort — but mid-effort against a brief that asks for more. fix: Prioritize executing the scale-tier brief over any cosmetic polish; the anchor + lever-scenarios are keepers, everything around them is the work. — RESOLVED 2026-06-10: spec executed (3 named readings + system angles), MockUI entities swept, EARN_NOTCH retuned ($800→+$100 capped 1/qtr; debt-first beats earn-first by ~$11K for the 22% persona, late-game exploit killed), riverstones→topical photography, NetWorthToolResearchProject replaces the form builder (flat persistence), fixtures repaired (regex targets translated; rig now normalizes them) — climb sim + shock proven in recapture, 0 interaction errors. ✓ resolved