Otterchip Model E — independent initial review (phase 1, corrected r2)

2026-11-05 · Claude Opus 5.5 · read-only audit · data: review-initial.json

The body model is a stature-scaled reference. Its segment ratios are age-interpolated historical estimates, and the default stature is the midpoint of the CDC 2000 sex-specific medians. Nothing here is a pediatric clinical optimum. The sources reviewed give no numeric child viewing or posture threshold. Bands are adult guidance or designer heuristics. No selected E design is assessed yet.

verified = code-verified, or verified on current M by the native real-adapter audit (native-geometry-audit.cjs/.json, 288 point checks, 0 failures). mechanism probe = my runs on a synthetic flat deck with a low pad and an assumed music point. These illustrate mechanisms only and are not current-M or E metrics. speculation = unverified.

Correction status (r2)

#Correction
1E is a fixed shell. Tilt is limited by the shell, display module and component envelope. The earlier hinge/stand claim is removed.
2Probe values are relabelled as mechanism probes. Current-M discontinuities now come from the native audit. Nothing here claims a neutral posture is unreachable.
3Reach now has two levels. The primary gate is home/representative pose. Far keys and controls are secondary tests with repositioning allowed. The 0.95 ratio is a heuristic flag.
4Auto furniture (same algorithm for every model) is the primary fair policy. Fixed furniture is optional sensitivity. Unverified furniture-class citations are removed.
5Non-government sources are paraphrased.
6Threshold wording is scoped to the reviewed sources. OSHA incidence guidance is verified.
7The music bridge uses native geometry. Collision is not assumed, and a geometric check is recommended.

1. Sources

2. Findings

V4 · Solver discontinuities on current M verified (high)

Native real-adapter audit, current Model M, age 6:

My mechanism probes show the same 25 mm hip-grid pattern. Required: score a ±2 mm neighbourhood, or refine hip position continuously in the optimizer wrapper. Report local sensitivity and re-solve rounded parameters.

V2 · No incidence metric exported by the solver verified (high)

screenNormal is read in (line 275) but never scored. The native audit computes incidence itself. The optimizer must report incidence at the screen centre and edges for each age, stature and activity. On E, tilt is fixed by the shell, display and component envelope.

V3 · Correlated viewing terms verified (high)

Head pitch = clamp(gaze − 20°), so neck flexion, eye-in-head angle, gaze, distance and incidence all come from one eye→screen geometry. Score them as one capped group.

V1 · Key-height stack tradeoff mechanism verifiedmagnitudes from probes (high)

When knee clearance sets the auto table height, as it did in every probe, key-top height above the table maps almost 1:1 into “keys above neutral”. In the synthetic probe, 33 mm lower keys gave about 32 mm less keys-above-neutral and 4–7° less forearm rise, with the table unchanged. Table thickness acts the same way. Within the searched assumptions, deck-stack height is the main lever, and screen tilt and setback trade against it. Judge E's actual result from its real adapter.

V5 · Reach scope verified in code (medium)

The solver tests reach only at the home-row target. Required, in two levels:

V6 · Music fingertip vs bridge — check needed gap verified (medium)

Native geometry:

Knuckles lie forward of the target, so a collision is not assumed. Before any fit claim, model:

Treat penetration over 1 mm as infeasible.

V7 · Furniture policy (low)

Use auto furniture, with the same algorithm, as the primary policy for M/D/E/W. A fixed household set is optional sensitivity.

V8 · Math correctness verified (medium)

The neutral-table sign is correct: ideal = seat + elbowAboveSeat − fingertipDrop − keyY, where positive means the keys are above elbow level. The angle definitions match their labels, and the native audit's 288 point checks passed. Score forearm pitch asymmetrically, with rising worse than slightly falling.

V9 · Component floors (medium)

Minimum deck = drop + max(LPK25 reserve + jack stack + 6, wheel stack). The LPK25 reserve of 26–36 mm is photo-estimated. Moving the jacks to the side gives a provisional 55 mm deck, with 57–59 mm recommended. Measure the LPK25 and check plug fit before release.

3. Stature sensitivity (CDC, mm)

ageP5 girlsex-median midpointP95 boy
38789471019
6106911531239
9123213341441
12139215041619

This is a sensitivity range only; segment proportions stay at the age preset.

4. Objective requirements

  1. Primary-pose hard feasibility. Require reach, no arm penetration over 1 mm, leg clearance, supported feet, and bridge clearance once checked. Weights cannot buy these back. Far-key and control reach are secondary.
  2. Ages. Use uniform weights for ages 3..12 and report the worst age beside the mean.
  3. Statures. Run P5 / midpoint / P95 (or ±10%) as sensitivity.
  4. Activities. The 70/30 split applies only to M and E. D and W get typing-only scores with no imputed music.
  5. Normalisation. Make all terms unitless. Show the adult 50–100 cm band as reference only.
  6. Grouping. Use capped viewing and arm groups. Show raw metrics first and the composite last.
  7. Weight sensitivity. Vary group weights ±50% and test 50/50 and 90/10 splits. Flag the result if the choice is weight-driven.
  8. Static-pose caveat. Lateral octave reach, wrist deviation, pronation and movement are not modelled. The pad is an obstacle, not a wrist target.

5. Solver-exploitation pitfalls

6. Speculation

speculation A 55–59 mm deck depends on measuring the LPK25 and on side-jack fit. speculation Probe magnitudes are illustrative only. speculation Real children's adaptations and comfort were not observed.

Phase 2 (screenshots, design, CAD) is pending.