Vantasner Hydride–Piezo
[DOCS]
01 PITCH 02 PROGRAM 03 DETAIL 04 SOURCE

[05 / Plan]

An empty villa to a running device in twelve months.

Target start: Monday 2 November 2026 (M1)

Three loops run in parallel every month: the lab builds and measures, a 24/7 agent fleet simulates ahead of the bench, and the team ramps to pay for both. Every month ends with a named decision.

[The promise] Ten years of research in twelve months. How the engine does it →

Open the roadmap The ask

[M1 · Mon 2026-11-02 · target]

An empty villa.

5 people · 4 servers · every long-lead order placed on day one

M1–M3FoundationsM4–M6Composition matrix and carrier decisionM7–M9Full-cell commissioning and as-built modelsM10–M12Signal-class predictions, design lock, first device campaign Shape of the year, pitch outline slide 25 · one milestone at every month end

[M12 · October 2027 · target]

An integrated device prototype running 300+ hours.

10 people · 16 servers · wall-plug heat balance at 1 Hz, full detector panel live

[05.1 · The roadmap]

The whole year,
one month at a time.

  • [Lab] Builds and measures four prototype generations in the villa laboratory.
  • [Engine] A 24/7 agent-run simulation queue narrows next month's lab parameters.
  • [Team + budget] Ramps from 5 to 10 people and pays for both.
  • [Decision] Every month closes one named decision: the bottom row of the chart.
MonthNov 2026 to Oct 2027
M1 Nov 2026 M2 Dec M3 Jan 2027 M4 Feb M5 Mar M6 Apr M7 May M8 Jun M9 Jul M10 Aug M11 Sep M12 Oct
PrototypeGen 0 to Gen 3
Gen 0 prepOrders, drawings, recipes
Gen 0 · M3–M5 Material-screening cell
Gen 1 · M5–M8 Calorimeter cell
Gen 2 · M8–M11 Piezo-hydride reactor
Gen 3 · M11–M12 Integrated device prototype
LabCommissioning order
Orders
Fit-out
Commissioning
First deuterium
Steady state
ComputeServers, 16 slots each
A 4 servers · 64 slots
B 8 servers · 128 slots
C 16 servers · 256 slots
TeamPeople, new hires lit
5
7
9
10
10
10
10
10
10
10
10
10
SpendLab capitalOperatingCHF, est.
238k
241k
287k
273k
245k
245k
191k
191k
232k
232k
232k
232k
DecisionClosed at month end
30 Nov Hydride shortlist order for stage 1
31 Dec Loading temperature windows
31 Jan Drive amplitude ceiling
28 Feb Piezoelectric composition ranking for Gen 1
31 Mar Carrier choice
30 Apr Loading ramp protocol adopted
31 May Duty cycle choice
30 Jun Final drive frequency
31 Jul As-built resonance correction
31 Aug Detector exposure plan
30 Sep Gen 3 frozen design
31 Oct Month-13 parameter set

Click any column or use the arrow keys on the month row. Dates are targets; costs and counts are planning estimates.

M1 November 2026 target

Procurement, queue commissioning, hiring

Decision closed Hydride shortlist order for stage 1

  • Prototype Gen 0 preparation
  • Compute Scale A · 4 servers · 64 slots
  • Team 5 people +5 new
  • Spend CHF 238,230 est.

[Build or modify]

sign the villa fit-out contractor (gas-room ventilation, dedicated circuits, gas piping, interlock wiring, flare line to roof); place all long-lead purchase orders on day one, per the LAB-BOM lead-time notes: CAEN DT5730S digitizer, Hiden HPR-20 QIC mass spectrometer, Pfeiffer PrismaPlus residual gas analyzer, MBraun LABstar glovebox, Parr Series 4520 reactor, Ludlum helium-3 neutron counter, Carbolite Gero split tube furnace; stand up the Marbell queue head server plus three workers; hire the first five staff.

[Run]

Marbell Scale A campaign, month 1 of the queue document: queue commissioning, density functional theory (DFT) foundations for PdD0.7, TiD2, ZrD2 and NiHx, screening and tunneling kernels v1, first ultrasonic horn eigenfrequency model.

[Measure]

DFT anchor set against literature (face-centered-cubic palladium lattice constant near 3.89 angstrom, beta-PdD expansion of 3 to 4 percent on full loading, fluorite TiH2 and ZrH2 stability, bulk moduli); horn first resonance prediction; first accepted-job rows in the results database with provenance hashes.

[Decision closed]

hydride shortlist order for stage 1 (TiD2 and ZrD2 primary platforms, PdDx mobility reference) and the horn geometry baseline.

Milestone · 2026-11-30 target

Seven long-lead purchase orders signed; Marbell Scale A queue running; the results database holds accepted DFT rows with the palladium lattice constant within 1 percent of 3.89 angstrom and the beta-PdD expansion inside the 3 to 4 percent reported band

Simulation queue · month 1

Queue commissioning; DFT foundations for PdD0.7, TiD2, ZrD2, NiHx; screening kernels v1; first horn model.

Narrows: Hydride shortlist for stage 1; DFT anchor set; horn first resonance

Jobs: 6,120 on the queue document's Scale B basis est.; this month runs at Scale A (4 servers), adjusted per that document's Scale A rules.

Month spend · CHF est.

  • Lab capital136,700
  • Consumables6,500
  • Compute6,000
  • Tokens12,800
  • Salaries67,000
  • Contingency accrual9,230
  • Month total238,230

M2 December 2026 target

Medium-lead orders, strain and phonon campaigns

Decision closed Loading temperature windows

  • Prototype Gen 0 preparation
  • Compute Scale A · 4 servers · 64 slots
  • Team 7 people +2 new
  • Spend CHF 240,630 est.

[Build or modify]

place medium-lead orders (pumps, gauges, Swagelok VCR manifold kit, Alicat mass-flow controllers, Bronkhorst Coriolis meter, Julabo circulator, Draeger transmitters); Gas Room fit-out continues first in the commissioning order; freeze the Gen 0 screening-cell drawings and the Parr-vessel loading recipes; hire the calorimetry lead and the materials engineer.

[Run]

Marbell Scale A, month 2 campaign: strain scans and phonons around the hydride transitions (200 K4, 400 K3, 60 K1, 8,000 K0 on the Scale B basis, reduced per the Scale A adjustments of the queue document), machine-learned interatomic potential (MLIP) benchmark phase, screening-kernel fitting v1.

[Measure]

temperature window per hydride from the quasi-harmonic expansion; thermal-expansion mismatch strain between hydride grain, piezoelectric grain and carrier; kernel benchmark against the measured metallic screening potentials of a few hundred electronvolts (Czerski 2001, Raiola 2004).

[Decision closed]

loading temperature windows (Pd 80 to 150 C, Ti 300 to 500 C, Zr 400 to 600 C) and particle-coating and cell-fit mismatch tolerances.

Milestone · 2026-12-31 target

Screening kernel v1 reproduces the consolidated metallic enhancement curves of Huke 2008 within a factor of 2 across the 0.1 to 10 keV grid (tolerance is a program target); loading temperature windows fixed for all three hydrides

Simulation queue · month 2

Strain scans and phonons around the hydride transitions; potential benchmark phase; screening fit v1.

Narrows: Temperature window; thermal expansion mismatch; kernel benchmark against measured metallic screening potentials

Jobs: 8,660 on the queue document's Scale B basis est.; this month runs at Scale A (4 servers), adjusted per that document's Scale A rules.

Month spend · CHF est.

  • Lab capital109,400
  • Consumables6,500
  • Compute6,000
  • Tokens12,800
  • Salaries94,000
  • Contingency accrual11,930
  • Month total240,630

M3 January 2027 target

Gas Room live, first powder in

Decision closed Drive amplitude ceiling

  • Prototype Gen 0
  • Compute Scale B · 8 servers · 128 slots
  • Team 9 people +2 new
  • Spend CHF 286,720 est.

[Build or modify]

place short-lead orders with the fit-out contractor on site (Omega sensors and controllers, NI modules, bench instruments, personal protective equipment); commission the Gas Room: leak-back check against 1e-7 mbar x L/s class sensitivity, PrismaPlus purity runs, flare and interlock walk-through; install and passivate the glovebox and the split tube furnace in the Workshop; receive first powder deliveries (titanium sponge and mesh powders, zirconium powder, palladium powder, barium titanate, PZT discs for crushing, fumed silica, deuterated carriers, isotope packs); hire the electrochemist and the acoustics and radio-frequency (RF) engineer.

[Run]

Marbell steps to Scale B: month 3 campaign, transient molecular dynamics (MD) ensembles (300 K1), reactive-field Ti-H and Zr-H cells (100 K2), discrete element method (DEM) force-chain packing v1 (150 K11), FEniCSx piezo-grain kernel v1 (800 K10). In the lab: first Parr-vessel deuteriding runs on titanium sponge chunks and -100 mesh powder, 1 to 10 bar D2, 300 to 500 C ramp-and-soak programs.

[Measure]

gravimetric D/Ti uptake per batch (target toward 1.9); sieved particle size distributions; rheology of first fumed-silica-in-polypropylene-glycol and deuterated-carrier suspensions (viscosity rise versus shear rate).

[Decision closed]

drive amplitude ceiling and the 1 to 10 micrometer particle size window for suspensions.

Milestone · 2027-01-31 target

Gas Room passes leak-back at or below the 1e-7 mbar x L/s class; first in-house TiD2 batch measured at D/Ti of at least 1.7 by gravimetric mass uptake; particle size window fixed at 1 to 10 micrometers

Simulation queue · month 3

Transient MD ensembles; DEM force chains v1; FEniCSx piezo grain kernel v1.

Narrows: Drive amplitude ceiling; particle size window; contact-stress pulse distributions

Jobs: 1,350 on the queue document's Scale B basis est..

Month spend · CHF est.

  • Lab capital109,400
  • Consumables3,600
  • Compute12,000
  • Tokens25,600
  • Salaries120,000
  • Contingency accrual16,120
  • Month total286,720

M4 February 2027 target

Detector baselines, four-composition screening matrix

Decision closed Piezoelectric composition ranking for Gen 1

  • Prototype Gen 0
  • Compute Scale B · 8 servers · 128 slots
  • Team 10 people +1 new
  • Spend CHF 272,740 est.

[Build or modify]

assemble the Seebeck-envelope calorimeter in the Cell Room (400-junction thermopile winding, isothermal enclosure, lagging); fix the detector mounts on the 1.0 m and 2.0 m floor rings; begin the two-week detector background campaign, full geometry, excitation off, exactly as the run-time geometry; pole and qualify the first piezoelectric powder lots (corona or contact poling at 3 to 5 kV/mm near 120 to 150 C, Berlincourt d33 recorded per lot).

[Run]

Gen 0 screening matrix in the quartz cells: four piezoelectric compositions (lead zirconate titanate PZT-5H and barium titanate as the active arms; zinc oxide and lithium niobate as the low-field comparison compositions) crossed with TiD2 and ZrD2 powders in deuterated oil shear-thickening fluid (STF), driven by the ultrasonic bath and the horn. Marbell Scale B, month 4: concept matrix scan (20,000 K0, 400 K9, 200 K11, 300 K3) over 4 hydrides x 4 piezoelectrics x carrier shortlist.

[Measure]

background count baselines per detector channel at fixed geometry; charge-signal response of sheared piezo-hydride suspensions against the electronic floor; per-lot d33 values.

[Decision closed]

piezoelectric composition ranking for Gen 1 and the carrier shortlist (dry bed versus deuterated oil versus heavy-water electrolyte).

Milestone · 2027-02-28 target

Two-week detector background campaign complete with per-channel baselines recorded at run geometry; all four piezoelectric lots poled with per-lot Berlincourt d33 on file; charge response of sheared piezo-hydride suspensions measured above the electronic floor

Simulation queue · month 4

Concept matrix scans: 4 hydrides x 4 piezoelectrics x carrier shortlist.

Narrows: Piezoelectric composition; volume fraction; carrier shortlist (dry bed versus deuterated oil versus electrolyte)

Jobs: 20,900 on the queue document's Scale B basis est..

Month spend · CHF est.

  • Lab capital82,000
  • Consumables1,800
  • Compute12,000
  • Tokens25,600
  • Salaries134,000
  • Contingency accrual17,340
  • Month total272,740

M5 March 2027 target

Commissioning complete, carrier choice

Decision closed Carrier choice

  • Prototype Gen 0 + Gen 1
  • Compute Scale B · 8 servers · 128 slots
  • Team 10 people
  • Spend CHF 245,440 est.

[Build or modify]

complete the 90-day commissioning order of LAB-BUILD section 9: Joule calibration of the Seebeck envelope at six setpoints from 0.1 to 20 W with the precision shunt kit; flow-calorimeter calibration (Coriolis meter plus matched Pt100 pairs); pickup census with all excitations firing into a dummy load while detectors count; mount the first stainless suspension cell into the envelope; hire the nuclear instrumentation lead (headcount reaches 10).

[Run]

first hydrogen-analog suspension runs end-to-end through every channel (calorimeters, neutron pair, charged-particle window, mass spectrometry), interleaved control mode exactly as the run cards prescribe. Marbell Scale B, month 5: carrier decision support (1,500 K10, 300 K9, 10,000 K0): field-survival arithmetic, Debye screening lengths, loaded acoustic models.

[Measure]

envelope conductance U and heat capacity from the Joule fits; noise floor after 1000 s averaging (design target at or below 50 mW); pickup counts against the twice-quiet-room-baseline limit; gap-field survival budget per carrier against the 1e8 to 1e9 V/m design target.

[Decision closed]

carrier choice. Expected outcome per the SPECS design lesson: deuterated oil or glycerol shear-thickening suspension first, heavy-water electrolyte cell as stage two driven by ultrasound and discharge transients.

Milestone · 2027-03-31 target

Seebeck envelope Joule-calibrated at six setpoints from 0.1 to 20 W with heat recovery at or above 99 percent (program target) and noise floor at or below 50 mW after 1000 s; pickup census at or below twice the quiet-room baseline; hydrogen-analog suspension cycle runs end-to-end through every channel

Simulation queue · month 5

Carrier decision support: field-survival arithmetic, Debye screening, loaded acoustic models.

Narrows: Carrier choice; gap-field budget feasibility; coating and fit tolerances

Jobs: 11,800 on the queue document's Scale B basis est..

Month spend · CHF est.

  • Lab capital54,700
  • Consumables1,800
  • Compute12,000
  • Tokens25,600
  • Salaries134,000
  • Contingency accrual17,340
  • Month total245,440

M6 April 2027 target

Gen 1 first light with deuterium

Decision closed Loading ramp protocol adopted

  • Prototype Gen 1
  • Compute Scale B · 8 servers · 128 slots
  • Team 10 people
  • Spend CHF 245,383 est.

[Build or modify]

Gen 1 calorimeter cell declared operational: stainless suspension cell with sonotrode dip tube, sapphire viewports, headspace line to the Hiden mass spectrometer, getters on the recirculation loop; run-card automation live in the DAQ; draft the first provisional patent filing on the drive scheme.

[Run]

first deuterated Gen 1 matrix entries: gas-loaded PdDx powder (D/Pd 0.70 plus or minus 0.05, gravimetrically confirmed) and TiD2 powder in the chosen carrier, driven at partial duty cycle, each run bracketed within 48 hours by randomized hydrogen-analog, no-piezo, no-hydride and excitation-off controls, with the five-sigma analysis threshold pre-registered in the run file. Per the LAB-BUILD commissioning order, the first deuterated dry-bed run enters the calendar only after a clean hydrogen-analog cycle. Marbell Scale B, month 6: MLIP dataset generation and first training (60 K6, 20 K12), 2D phase-field fronts (400 K7).

[Measure]

continuous 1 Hz heat balance from both calorimeters with input power measured at the cell feedthrough; loading monitor trace across the run; headspace mass spectra.

[Decision closed]

loading ramp protocol adopted, and the resistance-ratio style loading monitor made mandatory in every hydride stage.

Milestone · 2027-04-30 target

First deuterated Gen 1 run logged end-to-end with its randomized control bracket in the same run file; loading monitor reads D/Pd 0.70 plus or minus 0.05 on the reference powder; carrier decision documented with the gap-field budget per carrier

Simulation queue · month 6

MLIP dataset generation and first training; 2D phase-field fronts.

Narrows: Interface mobilities; ReaxFF replacement for shock cells

Jobs: 480 on the queue document's Scale B basis est..

Month spend · CHF est.

  • Lab capital54,643
  • Consumables1,800
  • Compute12,000
  • Tokens25,600
  • Salaries134,000
  • Contingency accrual17,340
  • Month total245,383

M7 May 2027 target

Duty cycle and phase fronts

Decision closed Duty cycle choice

  • Prototype Gen 1
  • Compute Scale B · 8 servers · 128 slots
  • Team 10 people
  • Spend CHF 190,740 est.

[Build or modify]

add the programmable perturbation-cycle feature to the drive electronics (small negative-then-positive input step at unchanged average power, per the Tohoku 2024 lesson); implement the saturation-first protocol in the run cards: every run logs a documented chemical absorption first stage before any post-saturation window is analyzed.

[Run]

Gen 1 matrix campaign on deuterated-glycerol and deuterated-alkane STF platforms with TiD2, ZrD2 and PdDx plus PZT-5H; drive-on versus drive-off interleave across the duty-cycle ladder. Marbell Scale B, month 7: 3D phase field under oscillatory strain (100 K8), full-cell coupled finite element method (FEM) models (200 K9, 200 K7).

[Measure]

heat balance per duty-cycle step; resonance drift versus temperature and suspension state from the network-analyzer impedance sweeps; phase-front timescales against the drive period from the 3D phase-field outputs.

[Decision closed]

duty cycle choice and the resonance-tracking control band versus temperature.

Milestone · 2027-05-31 target

Duty cycle chosen from calorimeter-balanced drive-on and drive-off data; resonance-tracking band versus temperature documented from impedance sweeps

Simulation queue · month 7

3D phase field under oscillatory strain; full-cell coupled FEM.

Narrows: Loading ramp protocol; duty cycle; resonance tracking versus temperature

Jobs: 500 on the queue document's Scale B basis est..

Month spend · CHF est.

  • Lab capital0
  • Consumables1,800
  • Compute12,000
  • Tokens25,600
  • Salaries134,000
  • Contingency accrual17,340
  • Month total190,740

M8 June 2027 target

Drive frequency locked; Gen 2 build starts

Decision closed Final drive frequency

  • Prototype Gen 1 + Gen 2
  • Compute Scale B · 8 servers · 128 slots
  • Team 10 people
  • Spend CHF 190,740 est.

[Build or modify]

build the 40 kHz comparison path (distributed wall transducers of the cleaning-transducer class bolted to a test vessel) alongside the 20 to 24 kHz horn; tune the L/C matching network to a voltage standing-wave ratio (VSWR) below 1.5 with the Bode 100; start the Gen 2 dry-bed reactor build: quick-open clamp cells with powder cartridge baskets, clamp-mounted piezo stack, pulsed-discharge bank interlocks per LAB-BUILD 6.2 (door-interlocked dump, armed stack light, fiber-optic trigger).

[Run]

frequency decision experiments: 20 kHz class horn versus 40 kHz distributed drive at equal absorbed power, with the camera watching sonoluminescence onset so the power ceiling is set below cavitation. Marbell Scale B, month 8: drive frequency and duty-cycle sweeps (500 K9, 80 K8, 200 K1).

[Measure]

absorbed acoustic power versus reflected power per frequency; cavitation onset amplitude per configuration; calorimeter-balanced drive-on and drive-off windows.

[Decision closed]

final drive frequency (20 versus 40 kHz class) and the input power ceiling below cavitation onset.

Milestone · 2027-06-30 target

Drive frequency locked with the 20 versus 40 kHz comparison on file, VSWR below 1.5 on the working network, and the power ceiling set below the optically observed cavitation onset

Simulation queue · month 8

Drive frequency and duty-cycle sweeps.

Narrows: Drive frequency final choice; power ceiling; duty cycle

Jobs: 780 on the queue document's Scale B basis est..

Month spend · CHF est.

  • Lab capital0
  • Consumables1,800
  • Compute12,000
  • Tokens25,600
  • Salaries134,000
  • Contingency accrual17,340
  • Month total190,740

M9 July 2027 target

Gen 2 bring-up; compute steps to Scale C

Decision closed As-built resonance correction

  • Prototype Gen 2
  • Compute Scale C · 16 servers · 256 slots
  • Team 10 people
  • Spend CHF 232,100 est.

[Build or modify]

complete the first dry-bed reactor: powder cartridge loaded in the glovebox (50 volume percent deuteride at 1 to 10 micrometers, 10 volume percent poled piezo powder, balance deuterated carrier or dry argon headspace), sealed, leak-checked, mounted with headspace getter loop and helium sampling line; add high-bandwidth pressure sensing plus multipoint temperature logging for the burst channels.

[Run]

Gen 2 first runs with hydrogen analogs and zirconia-bead blank beds, copying the Kobe and Tohoku oil mass-flow calorimetry method: chamber ambient at 23 plus or minus 0.1 C, blanks fixing the heat-recovery rate before any active bed runs. Marbell steps to Scale C, month 9: as-built cell models from measured geometry (600 K9, 2,000 K10, 5,000 K0 on the Scale B row, raised per the Scale C adjustments), nightly assimilation lane begins ingesting lab diagnostics.

[Measure]

as-built resonance against the Elmer prediction band; expected contact charge per drive amplitude; thermal balance of the blank bed; a deliberate perturbation cycle captured on the burst channels to prove the chain.

[Decision closed]

as-built resonance correction and the dry-bed drive power window.

Milestone · 2027-07-31 target

First Gen 2 dry-bed reactor logging heat balance with zirconia-bead blanks fixing the recovery rate at 23 plus or minus 0.1 C ambient; as-built resonance inside the Elmer prediction band; burst channels proven on a captured perturbation cycle

Simulation queue · month 9

Lab bring-up support: as-built cell models from measured geometry.

Narrows: As-built resonance prediction; expected contact charge; thermal balance

Jobs: 7,600 on the queue document's Scale B basis est.; this month runs at Scale C (16 servers), adjusted per that document's Scale C rules.

Month spend · CHF est.

  • Lab capital0
  • Consumables1,800
  • Compute24,000
  • Tokens51,200
  • Salaries134,000
  • Contingency accrual21,100
  • Month total232,100

M10 August 2027 target

Gen 2 isotope pairs and activation cycling

Decision closed Detector exposure plan

  • Prototype Gen 2
  • Compute Scale C · 16 servers · 256 slots
  • Team 10 people
  • Spend CHF 232,100 est.

[Build or modify]

second and third dry-bed cartridges (composition variants from the M4 ranking); re-oxidation-hydriding refresh station for powder lots per the Osaka lesson.

[Run]

matched D2 versus H2 pairs for at least two bed compositions, each with its documented saturation-first stage; activation cycling programs (load, bake, gas dose, repeat) scheduled with the 8 to 23 day-class activation windows of the SRI record; helium headspace protocol exercised end-to-end: getter strip of D2, mass-4 accumulation against the mass-20 internal reference, calibrated helium dose into an identical empty cell.

Run (compute): Marbell Scale C, month 10: signal-class kernels over the surviving matrix (30,000 K0, 50 K2, 20 K6): predicted proton line versus continuum spectra and minimum detectable yield targets.

[Measure]

per-pair heat balance over identical windows; CR-39 (polyallyl diglycol carbonate) track densities, open versus masked foils; neutron time-of-flight (ToF) coincidences on the 2.0 m baseline; helium accumulation rates.

[Decision closed]

detector exposure plan and the go-forward bed composition for Gen 3.

Milestone · 2027-08-31 target

Matched D2 and H2 pairs complete for at least two bed compositions with documented saturation-first stages; detector exposure plan with minimum detectable yield targets filed from the signal-class kernels

Simulation queue · month 10

Signal-class predictions: spectra kernels over the surviving matrix.

Narrows: Detector exposure plan; proton line versus continuum expectations; minimum detectable yield targets

Jobs: 30,070 on the queue document's Scale B basis est.; this month runs at Scale C (16 servers), adjusted per that document's Scale C rules.

Month spend · CHF est.

  • Lab capital0
  • Consumables1,800
  • Compute24,000
  • Tokens51,200
  • Salaries134,000
  • Contingency accrual21,100
  • Month total232,100

M11 September 2027 target

Design lock

Decision closed Gen 3 frozen design

  • Prototype Gen 2 + Gen 3
  • Compute Scale C · 16 servers · 256 slots
  • Team 10 people
  • Spend CHF 232,100 est.

[Build or modify]

integrate Gen 2 into the Gen 3 cabinet: reactor, ultrasonic and discharge drive electronics, both calorimeters, interlock chain, DAQ and control workstation in one package; wire wall-plug total calorimetry so input and output integrate from power-on to cool-down (the Mizuno 2019 air-flow method adapted to our envelope plus flow pair); file the second patent filing on cell geometry and the drive program.

[Run]

Marbell Scale C, month 11: design-lock sensitivity sweeps over the 12 design parameters (60 K8, 300 K9, 200 K3, 20,000 K0 on the Scale B row, raised per Scale C) with SALib Sobol analysis over the coupled model chain; final Gen 2 confirmation runs on the frozen bed composition.

[Measure]

Sobol sensitivity ranking and margin list against the measured Gen 2 dataset; full-parameter re-run of the winning composition.

[Decision closed]

Gen 3 frozen design and parameter brief v1.0 published from the results database.

Milestone · 2027-09-30 target

Parameter brief v1.0 with the sensitivity-ranked margin list over the 12 design parameters published from the results database; Gen 3 design frozen; second patent filing lodged

Simulation queue · month 11

Design lock: sensitivity sweeps over the 12 design parameters; margin list.

Narrows: Final parameter brief version 1.0; margin list for the lab

Jobs: 20,560 on the queue document's Scale B basis est.; this month runs at Scale C (16 servers), adjusted per that document's Scale C rules.

Month spend · CHF est.

  • Lab capital0
  • Consumables1,800
  • Compute24,000
  • Tokens51,200
  • Salaries134,000
  • Contingency accrual21,100
  • Month total232,100

M12 October 2027 target

Gen 3 integrated device prototype campaign

Decision closed Month-13 parameter set

  • Prototype Gen 3
  • Compute Scale C · 16 servers · 256 slots
  • Team 10 people
  • Spend CHF 232,100 est.

[Build or modify]

only corrective changes: the assimilation lane and the weekly re-planning cycle drive a short list of as-built corrections; nothing new is designed inside month 12.

[Run]

the Gen 3 integrated device prototype runs a continuous campaign of at least 300 hours (target): matched active and control reactors in the same thermal enclosure, wall-plug integration, programmable perturbation cycles on a fixed schedule, the full detector panel live throughout (moderated helium-3 counter, EJ-309 pulse-shape discrimination (PSD) pair on the ToF baseline, NaI(Tl) gamma channel behind the shadow bar, in-cell CR-39 pairs, silicon surface-barrier detector on the proton window, headspace helium mass spectrometry). Marbell Scale C, month 12: assimilation lane corrects the contact-stress and field models (1,000 K10, 200 K9, 10,000 K0) and feeds the next build cycle.

[Measure]

integrated input and output energy from wall plug to heat balance at 1 Hz; output-over-input ratio against the matched control; per-channel nuclear-product counts against the pre-registered five-sigma thresholds; helium accumulation against the calibrated blank.

[Decision closed]

the month-13 parameter set, the scale-up direction, and the program report to stakeholders with every result, either way, published with its full uncertainty budget.

Milestone · 2027-10-31 target

Gen 3 integrated device prototype completes a continuous campaign of at least 300 hours with wall-plug heat balance at 1 Hz and the full detector panel live against its matched control; program report issued with results, uncertainty budgets and the month-13 parameter set

Simulation queue · month 12

Campaign support: assimilate first device data, correct contact-stress and field models.

Narrows: Live parameter corrections feeding the next build cycle

Jobs: 11,200 on the queue document's Scale B basis est.; this month runs at Scale C (16 servers), adjusted per that document's Scale C rules.

Month spend · CHF est.

  • Lab capital0
  • Consumables1,800
  • Compute24,000
  • Tokens51,200
  • Salaries134,000
  • Contingency accrual21,100
  • Month total232,100

160

slot-years of round-the-clock agent work inside the twelve months est.

1,920 slot-months ÷ 12 · up to 256 slots

[05.2 · The promise]

Ten years of research in twelve months. The lab, a 24/7 agent fleet and the team run in parallel, and every month ends with a named decision.

How it works, in full, on the engine page.

Ten years in twelve months: how Parallel, not in sequence Which campaign feeds which decision

[05.3 · Four generations]

Four prototypes.
Each one narrows the next.

Every month's work is tied to one of four prototype generations. Each exists to close named design decisions, so the space left to search gets narrower with every build. Open a generation for what it is and what it decides.

Gen 0 Material-screening cellM3 to M5 target

[What it is]

Parr-class loading vessels for in-house deuteriding, quartz screening cells, ultrasonic bath and horn excitation, rheology and charge-proxy readout, no full calorimetry.

[Decisions it closes]

  • Hydride batch quality (D/M ratios)
  • piezoelectric composition
  • particle size window
  • carrier shortlist

Months: M3 M4 M5

Gen 1 Calorimeter cellM5 to M8 target

[What it is]

The 2 L stainless suspension cell inside the Seebeck-envelope calorimeter with the flow calorimeter as second instrument, full diagnostics chain, interleaved controls.

[Decisions it closes]

  • Carrier choice
  • loading ramp protocol
  • duty cycle
  • drive frequency
  • power ceiling below cavitation onset

Months: M5 M6 M7 M8

Gen 2 Piezo-hydride reactorM8 to M11 target

[What it is]

Sealed dry-bed reactor with powder cartridge, clamp-mounted piezo stack, pulsed-discharge drive, headspace helium sampling port, zirconia-bead blank calorimetry.

[Decisions it closes]

  • Powder-bed drive physics
  • activation cycling
  • D2 versus H2 isotope pairs
  • detector exposure plan
  • design margins

Months: M8 M9 M10 M11

Gen 3 Integrated device prototypeM11 to M12 target

[What it is]

Gen 2 reactor plus drive electronics, calorimetry, control and data acquisition (DAQ) in one cabinet, wall-plug total calorimetry, 24x7 operation.

[Decisions it closes]

  • Integrated operation
  • wall-plug heat balance
  • month-13 parameter set and scale-up direction

Months: M11 M12

PLAN-12-MONTH §1 Prototype generationsThe device

[05.4 · Targets from the record]

The published record
sets the targets.

The parameters the months work against are carried from the published papers and results the program builds on, and from the materials specifications. Each one links to where it comes from.

  1. 01Loading, palladium

    D/Pd 0.70 plus or minus 0.05 from gas loading in-house; push mobility-matched cells toward the D/Pd above 0.9 band that the SRI record ties to excess-heat episodes; resistance-ratio style loading monitor live in every hydride stage

    Source lesson SRI flow-calorimetry line: loading is a first-class instrumented parameter Evidence · cluster 03

  2. 02Loading, titanium and zirconium

    D/Ti toward 1.9 at 300 to 500 C, D/Zr toward 2.0 at 400 to 600 C, 1 to 10 bar D2 (estimate); the cheap deuteron density path

    Source lesson SPECS.md section 1

  3. 03Loading, palladium vessels

    1 to 5 bar D2 at 80 to 150 C, gravimetric mass-uptake confirmation of every batch

    Source lesson SPECS.md PdDx recipe

  4. 04Bed operating window

    150 to 350 C for 100 g-class powder beds with oil mass-flow calorimetry at 23 plus or minus 0.1 C ambient and zirconia-bead blanks fixing heat recovery

    Source lesson Tohoku and Kobe (Technova) method Evidence · cluster 04

  5. 05Saturation-first protocol

    Documented chemical absorption first stage; post-saturation windows only are analyzed for excess

    Source lesson Kobe 2016 record: heat appears after metal-hydride absorption saturates Evidence · cluster 04

  6. 06Activation windows

    Budget 8 to 23 day-class activation times and multi-week activation cycling into run plans

    Source lesson SRI appearance times; Mizuno glow-discharge activation Evidence · cluster 03

  7. 07Surface refresh

    Specify oxidation state per lot; periodic re-oxidation-hydriding cycles to restore absorption energetics

    Source lesson Osaka record: re-oxidation restored hydride-formation energies of 1.5 to 2.3 eV per D Evidence · cluster 04

  8. 08Drive, mechanical

    20 kHz horn drive primary (Hielscher UP400St, 24 kHz), 40 kHz distributed wall-drive comparison; power ceiling below cavitation onset

    Source lesson SPECS.md section 5 design lesson

  9. 09Drive, electrical

    Pulsed-discharge bank (10 mF film capacitors, 1.5 kV, about 11 kJ) for the dry bed; programmable perturbation cycles, small negative-then-positive input steps at unchanged average power

    Source lesson Brillouin pulse-drive architecture; Tohoku 2024 perturbation result, about 38 kJ per shown cycle Evidence · cluster 03

  10. 10Burst channels

    High-bandwidth pressure plus multipoint temperature logging in every cell, not only slow calorimetry

    Source lesson Tohoku replicated coincident pressure-plus-temperature bursts Evidence · cluster 04

  11. 11Temperature scope

    1,100 C-class incandescent reactors (Parkhomov line) stay out of scope for the villa program; our windows are 80 to 600 C loading and 150 to 350 C operation

    Source lesson Cluster 04 record Evidence · cluster 04

[05.5 · Running the loops]

Who, what
and when.

The three ramps behind the roadmap: people, servers and the laboratory itself, plus the slack built into the schedule.

[Team]

10 roles, ramped over four months.

Fully loaded Swiss planning costs est.. Two-person rules apply from day one for cylinder changes, bank arming, hydrogen recovery and discharge runs.

PLAN-12-MONTH §7 Team

M1 Nov 2026

5people

  • Principal investigator

    Physics targets, publication and filing decisions, stakeholder report

    CHF 18,000 per month, loaded est.

  • Program manager

    Budget, procurement, contractor, schedule, run-card governance

    CHF 13,000 per month, loaded est.

  • Simulation lead

    Marbell queue, campaign planning, parameter briefs

    CHF 15,000 per month, loaded est.

  • Automation and software engineer

    DAQ, queue glue, run-card automation, dashboards

    CHF 13,000 per month, loaded est.

  • Laboratory technician

    Cell assembly, powder handling, etching, cylinder logistics

    CHF 8,000 per month, loaded est.

M2 Dec 2026

7+2

  • Calorimetry lead

    Seebeck envelope, flow calorimeter, Joule calibration, uncertainty budget

    CHF 14,000 per month, loaded est.

  • Materials engineer

    Powders, sieving, coatings, poling, STF formulation

    CHF 13,000 per month, loaded est.

M3 Jan 2027

9+2

  • Electrochemist

    Loading protocols, electrolyte stage-two cell, loading monitors

    CHF 13,000 per month, loaded est.

  • Acoustics and RF engineer

    Horn and stack drive, matching networks, pulsed-discharge bank, perturbation programs

    CHF 13,000 per month, loaded est.

M4 Feb 2027

10+1

  • Nuclear instrumentation lead

    Neutron and gamma spectroscopy, CR-39, silicon detector, helium mass spectrometry, background campaigns

    CHF 14,000 per month, loaded est.

[Compute]

4 servers, then 8, then 16.

The fleet steps with the program's phase, at the fixed queue-document rate of CHF 1,500 per server-month and 16 agent-plus-simulation slots per server. 1,920 slot-months, CHF 180,000 est. for the year; LLM tokens for the agents another CHF 384,000 est. at CHF 200 per slot-month.

PLAN-12-MONTH §4 ComputeThe engine

  1. A 4 servers

    M1 to M2 · 64 slots · CHF 6,000/month est.

    Campaign and what it narrows

    Campaign Commissioning; DFT foundations; screening kernels v1; horn model; strain and phonon scans; kernel benchmark

    Narrows Hydride shortlist; DFT anchor set; temperature window; expansion-mismatch tolerances; kernel benchmark

  2. B 8 servers

    M3 to M8 · 128 slots · CHF 12,000/month est.

    Campaign and what it narrows

    Campaign Transient MD and DEM; concept matrix scan; carrier decision support; MLIP training; 3D phase field; frequency and duty sweeps

    Narrows Drive amplitude ceiling; particle size window; piezoelectric composition and volume fraction (about 10 percent baseline); carrier choice; loading ramp; duty cycle; drive frequency; power ceiling

  3. C 16 servers

    M9 to M12 · 256 slots · CHF 24,000/month est.

    Campaign and what it narrows

    Campaign As-built cell models; signal-class predictions; design-lock sensitivity sweeps; nightly assimilation of lab data with weekly re-planning

    Narrows As-built resonance and contact charge; detector exposure plan and minimum detectable yields; parameter brief v1.0; live model corrections from device data

[Lab]

The laboratory comes up in a fixed order.

Driven by the bill-of-materials lead-time tiers: long-lead items ordered in month 1, first deuterium only after a clean hydrogen-analog cycle.

PLAN-12-MONTH §5 CommissioningThe labLAB-BUILD §9
  1. M1 to M3 Orders

    Long-lead items ordered M1 (digitizer, mass spectrometers, glovebox, Parr reactor, helium-3 counter, furnace); medium-lead items M2 (pumps, gauges, VCR kit, flow controllers, Coriolis meter, circulator, gas transmitters); short-lead items M3 with the contractor on site (sensors, NI modules, bench instruments, protective equipment)

  2. M1 to M3 Fit-out

    Gas Room first (ventilation, flare line, interlocks), then electrical and Cell Room structures, then Workshop (furnace, glovebox)

  3. M3 to M5 Commissioning

    The 90-day sequence: gas-room leak-back and purity; calorimeter Joule calibration and noise floor; two-week detector background campaigns at run geometry; pickup census; first hydrogen-analog suspension runs end-to-end

  4. M6 First deuterium

    First deuterated dry-bed run enters the calendar only after a clean hydrogen-analog cycle, per the commissioning order

  5. M7 to M12 Steady state

    Weekly Joule calibrations, quarterly badge exchanges, annual inspections; two calorimeters and the full detector panel live for every run

[Schedule buffers]

Slack, built in.

  • 01

    Two float weeks are reserved inside M10 and M11 for activation-cycle overruns, because the record shows 8 to 23 day-class and multi-week activation times.

  • 02

    The Marbell fleet can step up one scale on one month's notice because servers are billed monthly.

  • 03

    Every long-lead laboratory item is ordered in M1 with the BOM's lead-time tier as the tracking baseline.

[05.6 · What it costs]

CHF 2.84M est.

CHF 2,839,023 for the twelve months, about USD 3.12M at the planning rate of 1 CHF = 1.10 USD. Salaries, lab capital, compute and tokens, consumables and contingency.

See the ask
238kM1241kM2287kM3273kM4245kM5245kM6191kM7191kM8232kM9232kM10232kM11232kM12

[05.7 · Go deeper]

The full plan,
section by section.

Everything on this page comes from the 12-month plan, Version 1.0 of 2026-10-21. Read any section in full, or jump to a month.

[Every section of the plan]

1. Prototype generations2. Operating targets carried from the reported record3. Month-by-month plan4. Compute scale schedule5. Laboratory commissioning order and ramp6. Budget7. Team8. Milestones9. Commercial path10. What the 12 months buy

[Every month in full]

M1: November 2026M2: December 2026M3: January 2027M4: February 2027M5: March 2027M6: April 2027M7: May 2027M8: June 2027M9: July 2027M10: August 2027M11: September 2027M12: October 2027
PROGRAM The ask: budget, IP and commercial path PROGRAM The engine: the 24/7 simulation queue PROGRAM The lab: villa laboratory and diagnostics SOURCE Pitch outline §9 and the queue campaign plan The shape of the year; queue §6 has the campaign for 4, 8 and 16 servers