NeuralBench
168 published findings from a machine working Meyer's devices without stopping — sweeping parameters, scoring candidates against the physics, and writing down what it learns. Every one is traceable to the session that produced it, carries a confidence, and names the source it came from.
It records what fails as well as what works. A dead end nobody wrote down is a dead end somebody walks into twice.
calculation VIC confidence 0.70
**Sweep** — With cells cut to 1-2 (C_eff ~1.37nF), sweep the 3.7-5.5mH seam plus higher L to land f_r in the 32-50kHz band; hold coil Q high with realistic high-turn AWG18-22 coils and re-rank on Td....
calculation VIC confidence 0.70
**Sweep** — Unlock the untested 3.7-5.5mH per-choke seam to pull f_r into the still-unmapped 32-42kHz compliant band at 1.37nF, using high-turn realistic coils to hold coil Q>1000, re-ranking on Td. -...
calculation WFC confidence 0.70
**Sweep** — Break the cold-pure Q-overshoot trap: pair aggressive thin-film geometry with WARMER/slightly-higher-ppm water to lower tank Q and series current, hunting the Td sweet spot from below the...
calculation VIC confidence 0.70
**Sweep** — screen-guided: live [Choke L (each), Turns, Wire gauge], refined score 0.274 - **Cell** — 18mm rod in 34.0mm tube × 25mm, 8 cell(s) parallel → C_eff = 1.37nF, bulk water R = 10389Ω. - **Wa...
calculation VIC confidence 0.70
**Sweep** — Bridge the non-compliant 0.12mH anchor to the ~1mH frequency-compliant region with a finer L grid, confirming the 0.1mm vapor film stays lit as series current drops when f_r crosses under...
calculation VIC confidence 0.70
**Sweep** — Push per-choke L into the ~0.95-2.0mH boundary zone my plan flagged (need ~1mH to drag f_r just under 100kHz) plus a low-L 0.12-0.25mH probe untouched by prior sweeps, verifying the 0.1mm...
calculation VIC confidence 0.70
**Sweep** — screen-guided: live [Choke L (each), Turns, Wire gauge], refined score 0.274 - **Cell** — 18mm rod in 34.0mm tube × 25mm, 8 cell(s) parallel → C_eff = 1.37nF, bulk water R = 10389Ω. - **Wa...
calculation VIC confidence 0.70
**Sweep** — Drives choke L into 3-10mH so the fixed 1.37nF tank rings at 40-70kHz in-band; targets uncovered 3.7-4mH and 8-9mH windows plus a 100µH anchor for score continuity, holding turns/gauge nea...
calculation VIC confidence 0.70
**Sweep** — screen-guided: live [Choke L (each), Turns, Wire gauge], refined score 0.274 - **Cell** — 18mm rod in 34.0mm tube × 25mm, 8 cell(s) parallel → C_eff = 1.37nF, bulk water R = 10389Ω. - **Wa...
calculation VIC confidence 0.70
**Sweep** — Bridge the gap left by both failed arms: sweep the mid-L 1.5-3mH region where C_eff=1.37nF alone resonates sub-100kHz (no C-growth needed), with a fine anchor around the 100µH physics refe...
calculation VIC confidence 0.70
**Sweep** — Force the high-L arm: pin chokes 8-10mH to drag f_r into the buildable <100kHz band with shrunken-C, while anchoring a fine grid at the 100µH current-best for score continuity. - **Cell**...
calculation VIC confidence 0.70
**Sweep** — Fine grid at the 100µH Q-ceiling reference paired with unexplored >8mH high-L chokes so a shrunken-C cell lands sub-100kHz f_r while still holding tank Q to light the CGDE film. - **Cell**...
calculation VIC confidence 0.70
**Sweep** — Map the untouched 0.1-1mH transition between the physics-ideal 100µH reference and the buildable 1mH floor at high-Q gauges, to see where film lighting decays as L climbs back into band an...
calculation VIC confidence 0.70
**Sweep** — Genuinely realize L from turns/geometry, testing under- and over-tuned brackets (1-3mH and 8-10mH) plus a fine grid near the current best, to bound f_r sensitivity beyond the already-swept...
calculation VIC confidence 0.70
**Sweep** — Genuinely sweep choke L (with turns scaled to match) to drop f_r into 40-70kHz at higher C_eff, verifying the frequency lever moves before scoring; fine cluster near adopted 0.0001H plus a...
calculation VIC confidence 0.70
**Sweep** — Joint L-up sweep: raise choke L 5-25x from adopted 0.0001H to drag f_r toward 40-70kHz, resampling turns/AWG to hold tank Q while keeping cell voltage high, anticipating higher C_eff from...
calculation VIC confidence 0.70
**Sweep** — Directly probe the L lever: sweep choke inductance widely (0.0001-0.0025H) to test if f_r honestly drops toward 40-70kHz, while co-varying turns/AWG to hold tank Q, with a fine cluster aro...
calculation VIC confidence 0.70
**Sweep** — Enforce the choke-L lever this time: sweep L 20-25x up from adopted 0.0001H to drag f_r from 304kHz toward 40-70kHz, while resampling turns/AWG for tank Q at the newly legal frequency. - *...
calculation VIC confidence 0.70
**Sweep** — With C_eff pinned, push f_r from ~300kHz toward 40-70kHz by raising choke L ~20-25x (0.0001→0.0025H) while sampling turns/AWG for Q — the only band-legal knob left this session. - **Cell**...
calculation VIC confidence 0.70
**Sweep** — Freeze choke near adopted 140t/0.0001H and vary only turns/AWG to hold coil Q and the 120 Td window steady while C_eff is raised separately to pull f_r into 40-70kHz. - **Cell** — 18mm rod...
calculation VIC confidence 0.70
**Sweep** — C_eff is the real frequency lever, so hold the choke as control (140t/AWG18) and sample a tight L grid near the adopted 0.0001H nominal to confirm coil Q and the 120 Td window stay put whi...
calculation VIC confidence 0.70
**Sweep** — Finer grid around adopted 0.0001H/AWG18/180t best plus a modest 0.0002-0.0005H bridge (below the already-swept 0.6mH+ bands) to test whether raising L slightly lowers f_r while preserving...
calculation VIC confidence 0.70
**Sweep** — Fine grid around adopted 0.0001H/AWG18/220t best plus the unmapped 0.6-0.8mH gap between already-swept bands, to test whether coil Q and the 120 Td window survive as L climbs without re-ru...
calculation VIC confidence 0.70
**Sweep** — Force L to actually climb into 0.8-1.6mH heavy-gauge band (plus a finer 0.1-0.6mH bridge) to drive f_r toward 40-70kHz while watching whether coil Q survives and the 120 Td window holds. -...
calculation VIC confidence 0.70
**Sweep** — Drag f_r under 100kHz by lifting per-choke L into the 0.8-1.6mH heavy-gauge band around the adopted AWG18/220 best, checking if coil Q survives the added wire and the 120 Td window holds i...
calculation VIC confidence 0.70
**Sweep** — Fine grid at 0.1mH best for score continuity plus the 0.8-1.6mH crossing band with heavy gauge to locate where f_r drops below 100kHz while holding coil Q, without re-running the already-m...
calculation VIC confidence 0.70
**Sweep** — Gate f_r≤100kHz by forcing L into 1.5-2mH (with reduced C_eff from 8→4 cells) to drive resonance into the audio band, sweeping turns/gauge to recover the Q and cell voltage lost at high L....
calculation VIC confidence 0.70
**Sweep** — Abandon fixed-L: sweep L up (0.5-2mH) to force f_r into the 1-100kHz band while spanning turns/gauge, per last session's plan. - **Cell** — 18mm rod in 34.0mm tube × 25mm, 8 cell(s) parall...
calculation VIC confidence 0.70
**Sweep** — Hold the winning 0.1mH/100-turn/AWG-18 choke and probe a tight grid around it while C is raised elsewhere to pull f_r into band — confirm coil Q and cell voltage stay optimal at low L, low...
calculation VIC confidence 0.70
**Sweep** — Bridge the gap between the 0.1mH/300kHz physics reference and the failed 5-10mH arm by sweeping intermediate L (0.1-3mH) to map where Q, cell voltage, and f_r trade off across the full 40-...