{
  "title": "Otterchip Model E — independent initial review (phase 1: evidence and objective audit, corrected)",
  "date": "2026-11-05",
  "revision": "r2 — phase-1 corrections applied after native coordinator review (see correctionStatus)",
  "reviewer": "Claude Opus 5.5 (claude-opus-5-5), independent reviewer; writes only hardware/otterchip_model_e/review/**",
  "scope": "Read-only audit of web/otterchip-ergonomics-math.js, web/otterchip-ergonomics.js, web/otterchip-model-m-v6.js (musicMinimumDeck), web/otterchip-component-lpk25.js (fit reserve), hardware/otterchip_research/ergonomics/{sources.json,review-findings.html}, hardware/otterchip_model_e/{mechanical/coordination-note.txt,optimization/opus-brief.txt}, and the native real-adapter audit review/native-geometry-audit.{cjs,json}. No selected Model E design is assessed here.",
  "framing": "The modelled body is a stature-scaled reference built from age-interpolated, historically estimated segment ratios. Stature defaults to the arithmetic midpoint of the CDC 2000 male/female stature-for-age medians. No output of this solver or any optimizer is a pediatric clinical optimum. The sources reviewed here give no numeric pediatric viewing or posture health threshold; this review does not claim that none exists anywhere. All bands are adult guidance or designer heuristics.",
  "correctionStatus": [
    { "item": 1, "status": "corrected", "change": "E is a fixed shell with no hinge or monitor stand. Screen tilt is constrained by the fixed shell, the display module and the component envelope. The earlier hinge/stand wording is removed." },
    { "item": 2, "status": "corrected", "change": "My Node runs used a synthetic flat key-height deck, a low pad and an assumed music point. They are now labelled approximate mechanism probes, not current-M metrics. Current-model values come from the native real-adapter audit (288 point checks passed), and its discontinuities lead V4. V1 is reworded as a stack-height tradeoff within the searched assumptions; it no longer claims a neutral posture is unreachable or that all designs fail." },
    { "item": 3, "status": "corrected", "change": "Reach is split into two levels. The primary gate is home/representative-pose reach. Far keys and controls are secondary tests that allow plausible hand, arm and trunk repositioning. The 0.95 reach ratio is a heuristic flag, not a physical impossibility." },
    { "item": 4, "status": "corrected", "change": "Auto furniture (the same algorithm for every model) is the primary fair policy. Fixed furniture is an optional sensitivity check only. References to unverified child-furniture classes are removed." },
    { "item": 5, "status": "corrected", "change": "Non-government source verification is paraphrased, with no long verbatim quotations. OSHA quotations are kept short." },
    { "item": 6, "status": "corrected", "change": "Wording now says the reviewed sources give no numeric child viewing threshold. OSHA's monitor tilt/incidence guidance is marked verified." },
    { "item": 7, "status": "corrected", "change": "The music bridge finding now uses the native geometry: black-key-top front at local 15.6 mm and bridge edge ≈ underlap − 49.5 mm. A fingertip–ceiling collision is not assumed; a geometric check is recommended before any claim." }
  ],
  "methods": {
    "currentModelValues": "Native real-adapter independent audit: review/native-geometry-audit.cjs / .json, current installed Model M default, 288 point checks, 0 failures. This is the authority for current-M metrics.",
    "approximateMechanismProbes": "My own Node runs of the shipped solver used auto furniture, a 26 mm table and tableFrontZ = bodyFrontZ + 20. They also used a SYNTHETIC flat deck at key-top 101.8 mm (or 68.8 mm), a low palm pad and an ASSUMED music point (y 80, z 90). These runs illustrate solver mechanisms and sensitivities only. They are NOT current-M or E metrics and must not be tabulated as such.",
    "statures": "CDC statage.csv (https://www.cdc.gov/growthcharts/data/zscore/statage.csv) fetched 2026-11-05; P5 girl, sex-median midpoint and P95 boy read at Agemos = 12*age + 0.5.",
    "scratch": "Probe script kept outside the repo in /tmp/oe-review."
  },
  "sourceVerification": [
    { "id": "osha-monitors", "kind": "US government guidance (adult office)", "url": "https://www.osha.gov/etools/computer-workstations/components/monitors", "verified": "fetched 2026-11-05",
      "summary": "Preferred viewing distance 50–100 cm. Top of screen at or slightly below eye level. Screen centre normally 15–20° below horizontal eye level. Tilt the monitor 'perpendicular to your line of sight' (incidence), with a warning that tilting back can add ceiling glare.",
      "implication": "Verified basis for an incidence-angle metric. Distance and gaze bands are adult references, not child pass/fail rules." },
    { "id": "cornell-cuwe-children", "kind": "university expert guidance (paraphrased)", "url": "https://ergo.human.cornell.edu/cuweguideline.htm", "verified": "fetched 2026-11-05",
      "summary": "For children, the page advises:\n- an open elbow angle, with the forearm slightly below horizontal;\n- a near-neutral wrist (under about 15°);\n- feet supported on the floor or a footrest;\n- relaxed upper arms;\n- a negative-tilt keyboard;\n- a screen viewable without noticeably tipping the neck.\nIt gives no numeric screen distance, gaze angle or keyboard height.",
      "implication": "Qualitative child guidance. Supports asymmetric forearm scoring, where rising is worse than slightly falling." },
    { "id": "cdc-statage", "kind": "US government data", "url": "https://www.cdc.gov/growthcharts/data/zscore/statage.csv", "verified": "fetched 2026-11-05",
      "values_mm": { "3": { "P5girl": 878, "midP50": 947, "P95boy": 1019 }, "6": { "P5girl": 1069, "midP50": 1153, "P95boy": 1239 }, "9": { "P5girl": 1232, "midP50": 1334, "P95boy": 1441 }, "12": { "P5girl": 1392, "midP50": 1504, "P95boy": 1619 } },
      "implication": "Within-age stature spread (about ±7%) is comparable to the step between adjacent ages, so a stature-sensitivity pass is warranted." },
    { "id": "not-refetched", "note": "BAPAM, Yamaha and the other OSHA pages were not refetched in this bounded pass; see research/ergonomics/sources.json." }
  ],
  "findings": [
    { "id": "V1", "severity": "high", "evidenceLevel": "mechanism verified in solver code; magnitudes from approximate probes only",
      "title": "Within the searched assumptions, key-top height above the table is a strong, nearly 1:1 lever on forearm rise",
      "evidence": "Auto furniture places the table at the knee-clearance minimum whenever that exceeds the neutral-elbow table, which happened in every probe. Key height then passes almost directly into keyAboveNeutralMm. Approximate probe on a synthetic flat deck: dropping key top by 33 mm lowered keyAboveNeutral by about 32 mm at ages 3–12. Forearm pitch fell 4–7° and the auto table height did not change. Table thickness acts the same way.",
      "consequence": "Deck-stack reduction is the main physical lever. Screen tilt and setback trade against it but do not replace it. Whether a given E stack reaches a near-neutral forearm must be judged from E's real adapter and the optimizer's search, not from these probes. Report the remaining tradeoff honestly.",
      "requirement": "Report keyTopAboveTableMm, keyAboveNeutralMm and forearm pitch raw for every model and age before any composite." },
    { "id": "V2", "severity": "high", "evidenceLevel": "verified in code",
      "title": "The shipped solver does not export a screen incidence metric",
      "evidence": "screenNormal is sanitised (ergonomics-math.js:275) but no metric reads it. The native audit computes incidenceDeg independently.",
      "requirement": "The optimizer/report must compute incidence = angle between (eye − screen centre) and the screen normal from the solved eye point, per age, stature and activity, ideally at the screen centre and edges. On E, tilt is limited by the fixed shell, the display module and the component envelope; there is no hinge or stand." },
    { "id": "V3", "severity": "high", "evidenceLevel": "verified in code",
      "title": "Viewing terms are one correlated signal",
      "evidence": "Auto head pitch = clamp(gaze − 20°) (lines 559–567). neckFlexDeg, eyeInHeadDownDeg, gaze, distance and incidence all derive from one eye→screen geometry.",
      "requirement": "Score viewing as one grouped, capped term. Do not sum gaze, neck flexion and distance as independent rewards." },
    { "id": "V4", "severity": "high", "evidenceLevel": "verified on current M by native real-adapter audit; corroborated by approximate probes",
      "title": "Solver discontinuities from the coarse hip-position search",
      "evidence": "Current M geometry, age 6 (native audit):\n- deck 91.5 → 92 mm: hip 391.6 → 366.6 mm, elbow 95.5° → 88.7°;\n- deck 96 → 96.5 mm: hip 368.6 → 401.6 mm, elbow 88.6° → 98.2°.\nIndependent mechanism probes showed the same pattern: jumps of about 37 mm in hip position and a sawtooth with about 25 mm period when sweeping key z, matching the 25 mm coarse hipZ grid. The arm grid of 2.5° × 5° adds smaller steps.",
      "requirement": "The optimizer should either score a small neighbourhood (for example ±2 mm per geometric parameter) or refine hipZ continuously in an optimizer-side wrapper. It must report local sensitivity at the chosen point and re-solve rounded parameters rather than inheriting the numeric best." },
    { "id": "V5", "severity": "medium", "evidenceLevel": "verified in code (scope of reach test)",
      "title": "Reach is tested only at the home-row target",
      "evidence": "handTargets uses one keyPoint ± halfSpread. Humans reposition their hands, arms and trunk between keys, so far-key reach from one static pose is not a physical gate.",
      "requirement": "Use two levels:\n(1) PRIMARY GATE: home/representative-pose reach for typing and music. The fingertip miss must be within the existing heuristic, with no arm–shell or arm–table penetration over 1 mm.\n(2) SECONDARY TESTS: far keys (GK64X corners, LPK25 ends) and controls, allowing plausible repositioning (trunk lean or shift, shoulder flexion). Report the normalised reach ratio (shoulder→target ÷ upper arm + forearm + hand, from the solved acromion). A ratio around 0.95 is a heuristic flag for strain or repositioning, not impossibility. Secondary results inform the design but need not veto it." },
    { "id": "V6", "severity": "medium", "evidenceLevel": "partial-model gap verified in code; bridge geometry from native note; collision NOT established",
      "title": "Music fingertip vs typing-bed bridge needs an explicit geometric check",
      "evidence": "The solver checks only the tabletop, a deck plane/profile and the pad block. Native geometry:\n- black-key-top front is at local 15.6 mm;\n- the bridge edge is ≈ underlap − 49.5 mm;\n- exposed black-key top is about 26.1 mm at underlap 39 and about 20.1 mm at underlap 45.\nKnuckles lie forward of the target, so a ceiling collision is not assumed.",
      "requirement": "Before any fit claim, model:\n- black-key contact 5–10 mm behind the black-key front;\n- a fingertip capsule of at least 8–10 mm radius plus 3 mm clearance against the bridge underside;\n- the knuckle position from the solved hand.\nTreat penetration over 1 mm as infeasible. Until this check runs, report underlap as a mechanism tradeoff, not a collision." },
    { "id": "V7", "severity": "low", "evidenceLevel": "policy note",
      "title": "Furniture policy",
      "evidence": "Auto furniture applies the same algorithm to every model and is the requested fair primary policy. It does not by itself hide device benefit, because key and screen geometry still change arm and viewing metrics.",
      "requirement": "Use auto furniture as the primary policy for M/D/E/W. A fixed household table and seat is an optional sensitivity check, not a required standards search." },
    { "id": "V8", "severity": "medium", "evidenceLevel": "verified by hand derivation",
      "title": "Neutral-forearm table sign and angle definitions are correct",
      "evidence": "ideal = seat + elbowAboveSeat − fingertipDrop − keyY (line 688). Positive keyAboveNeutral means the keys are above relaxed elbow level. Elbow, wrist and gaze definitions match their labels. The native audit's 288 point checks found no failures.",
      "requirement": "Keep recomputing angles from exported points only. Score forearm pitch asymmetrically: rising is worse than slightly falling." },
    { "id": "V9", "severity": "medium", "evidenceLevel": "formula verified in code; component values partly photo-estimated",
      "title": "Physical floors on deck height come from components",
      "evidence": "musicMinimumDeck = drop + max(reserve + front-jack stack + 6, wheel stack). The LPK25 reserve range [26,36] mm is a photo estimate. Per the coordination note, moving the jacks to the left side gives a provisional 55 mm flat deck (57–59 mm recommended).",
      "requirement": "Use these as hard optimizer bounds. Release depends on measuring the LPK25 reserve and checking plug and housing fit at the side." }
  ],
  "objectiveRequirements": [
    "Hard feasibility for the primary seated pose:\n- home/representative reach;\n- no arm penetration over 1 mm against real E surfaces;\n- legs clear of the table underside;\n- feet supported (footrest allowed);\n- music fingertip clear of the bridge once V6 is checked.\nWeights cannot buy these back. Far-key and control reach are secondary tests with repositioning allowed.",
    "Ages: uniform 3..12 as the primary objective, with the worst age reported beside the mean. Treat population weighting as an optional variant.",
    "Stature sensitivity: CDC P5 girl / midpoint / P95 boy at each age (or the native ±10% scale). Label it as sensitivity, not percentile bodies.",
    "Activities: the 70/30 split applies only to M and E. D and W get a typing-only column and no imputed music score.",
    "Normalisation: every term unitless.\n- Angles: deviation ÷ heuristic band.\n- Reach: distance ÷ arm chain from the acromion.\n- Viewing distance: ratio to an age-scaled reference.\nShow the adult OSHA 50–100 cm band as a reference line only.",
    "Grouping: one capped viewing group (gaze, incidence, distance, neck) and one arm group (elbow, forearm, shoulder, wrist). Show raw metrics first and the composite last.",
    "Weight sensitivity: vary group weights ±50% and the activity split to 50/50 and 90/10. If the selection flips, state that the choice is weight-driven.",
    "Dynamic caveat: these are static single poses. Lateral octave reach, wrist deviation, pronation and continuous movement are not modelled. The palm pad is an obstacle, not a wrist target."
  ],
  "solverExploitationPitfalls": [
    "Band statuses are step functions, so use continuous deviations.",
    "Coarse hip and arm grids cause jumps (V4, verified on current M).",
    "Pushing the screen back trades distance against incidence and gaze. Pulling the keys forward trades reach against front-lip and bridge clearance. Only the geometric checks block these.",
    "Shell features missing from deckProfile are invisible to the solver. halfSpread is clamped to 0.2·bodyWidth.",
    "Seat rounding, 5 mm footrest steps and leg-branch switches add small discontinuities.",
    "Children's adult viewing metrics have status 'info' and are excluded from 'overall'. Keep viewing in the objective without making it a child pass/fail rule."
  ],
  "comparisonPolicy": {
    "models": ["M (default)", "D", "E", "W"],
    "furniture": "Primary: auto furniture, same algorithm for all. Optional sensitivity: one fixed household set.",
    "typing": "All four on identical ages, statures and metrics.",
    "music": "M and E only; D/W show n/a.",
    "display": "Raw per-metric age curves, then a feasibility table, then a grouped composite with a weight-sensitivity band."
  },
  "componentConstraints": [
    { "part": "GK64X typing PCB + caps", "constraint": "Full 19.05 mm pitch; about 16.6 mm below the deck; keycaps about 10.8 mm above the deck at M.", "status": "repo geometry" },
    { "part": "Akai LPK25 Mk2 keybed", "constraint": "Reserve below black-key top 26–36 mm; travel 6–10 mm. Together these set the minimum deck under the bridge.", "status": "photo-derived; measure before release" },
    { "part": "Headphone/line jacks", "constraint": "In M the front-jack stack sits under the keybed. Relocating it to the side removes that term, subject to plug body and cable bend clearance.", "status": "proposal; unverified" },
    { "part": "10.5\" AMOLED in fixed shell", "constraint": "Panel and frame set the screen-top height above the deck. Tilt is fixed by the shell, display module and component envelope (no hinge or stand). Weigh the incidence and glare tradeoff per OSHA.", "status": "dimensions in repo" },
    { "part": "User tabletop thickness", "constraint": "Raises keys above neutral roughly 1:1 when knee clearance sets the table height.", "status": "mechanism verified in solver" }
  ],
  "speculation": [
    "A 55–59 mm deck depends on the LPK25 reserve measurement and the side-jack fit.",
    "Probe magnitudes (synthetic flat deck, assumed music point) are illustrative only.",
    "Real children's adaptations and comfort were not observed."
  ],
  "phase2Pending": "Screenshot, design and CAD review will follow when the images and selected parameters arrive."
}
