Each priority in the supplied assessment was re-examined against its source material, then pushed forward with whatever the question actually needed: a reproduction, a correction, a numerical model, a measurement design, or a bill of materials.
Reordering whole days inside each month raises ideal storage capacity by 10.7 %.
Reproduced at 10.8 %. Falls to 5.1 % with 3-day blocks and 1.8 % with 7-day blocks; 1.3 % of least-cost system cost once persistence is preserved.
Strongest existing analysis — but the headline number is a property of the null
Next: Publish the null-sensitivity curve, not the single number.
A finite-cutoff model cannot represent the elastic energy outside its cutoff, so error ∝ P²/(κR_c³).
A 512-atom lattice gives exponent +0.25 in the natural comparison and −3.0 once the bulk modulus is matched — the sign follows the parameterisation, so R_c⁻³ is not a signature. The channel is identifiability, not missing energy.
Promising direction — the derivation needs correction before any benchmark
Next: Run the single-material, 30–50 defect DFT pilot.
Two swelling pathways with matched gain and different time constants give a rate signal.
Confirmed analytically: a band-pass differentiator peaking at 1.3 mHz for a 20/60 µm pair, detection floor 5×10⁻⁵ /s. "Rapid" means minutes.
Attractive — now a specified instrument with a detection floor and a €2.7k test
Next: Bench test: frozen-parameter prediction + matched-time control.
Interleaved inert witness patches estimate deposition bias better than edge-placed blanks.
True on every field tested at equal cost: 0.8–1.0 % vs 1.9–37.9 % error. For a field constant on the boundary the edge layout is wrong by 22.9 % at any N.
Clear mathematical result — now a design rule with a failure theorem
Next: One substrate batch with an independent full-field map settles it.
NV-diamond magnetometry detects the current constriction before a dendrite short.
A 10 mA filament gives 0.48 nT at 3 mm; an OPM sees it at SNR 3×10⁴, so sensor noise never binds. The 5 A charging background is 87 nT–30 µT, requiring common-mode rejection of 181–5 800 — cut to 3.9 by a 1 mm gradiometer with 4×10⁻⁴ balance. A 10-minute warning needs a growth constant τ ≳ 85 s.
High value — but the binding constraint is the charging field, not the sensor
Next: Wire-phantom bench test on an OPM/SQUID gradiometer.
Sparse perforations cut swelling delay more than their fractional open area predicts.
208 finite-volume layouts: the gain over a proper area-only control never exceeds 1 (0.084–0.961, median 0.595). The deficit is set by hole radius over skin thickness — G ≈ 1/(1+0.45(a/δ)^0.84), r = 0.898 — not by open area. At fixed open area, hole size changes t50 by 6.5×, and "sparse" is the worst configuration available.
Cheap, falsifiable — the disproportionate gain has a narrow window at best
Next: A single mass-uptake series on three hole densities would settle it.
Pulsed duty ages Na-ion cells 1.3× faster at matched RMS current.
At matched throughput pulsing lowers mean overpotential (×0.48–0.92) and raises the peak (×1.8–4.8). Plating is a duty threshold and a ~10⁻³ Hz window.
Practical and commercial — re-scoped after a 2026 negative study
Next: Map plating onset in (frequency, duty); 3 cells per point.
Tactile models give relative feel; nothing calibrates absolute newtons or stiffness.
Absolute force U(k=2) = 9.46 % (2.93 % with a matched loading history; 13.8 % as-shipped), stiffness 14–16 %, dominated by hysteresis and contact geometry, not by force realisation. The original kill criterion is false: hysteresis kills contrast below ~19 %, not absolute force. Phantom allowance 1.5 %, met at 1.42 % for €3,200.
Useful benchmark — the open claim is an uncertainty chain, not absolute newtons
Next: One hysteresis loop on one real sensor replaces the biggest assumption.