theFlux Project
Pilot Phase 1 — Fast-Track Action Plan
Three Months · $500,000 CAD · 1 fluxNode · 3 fluxPoles · 3 fluxPods · 4 Development Teams
Prepared for theFlux Corp — Icarus / John Argast
Rev A — July 18, 2026 · Working draft — internal planning document
1. Executive Summary
This plan compresses theFlux Pilot Phase 1 into a single 3-month Program Increment (six two-week sprints) on a fixed $500,000 CAD budget. It stands up one complete fluxNode — three fluxPoles, each carrying three Mac mini M4 units (nine compute nodes total) — co-located with a three-pod settlement cluster: one storage pod and two living/work pods, laid out on the campground-loop pattern. Four paid fractional contract leads each run one development team: AI, RF, Civic, and Democracy Level II. Teams are hybrid — small stipend cells drawing on open-source contributors and North Island College students, compensated through stipends, Class C option grants, and DOJO/dataset access.
The plan follows the Agile Action Plan Rev B operating model (two-week sprints, hard gates, Definition of Done discipline) but scopes claims honestly: one node is a node pilot, not Capacitor conformance — the Rev A spec requires five nodes and a full coastal winter before any conformance claim. Everything demonstrated here (mesh quorum, layered spectrum, federated learning, one instrumented deliberation) is a gate on the road to that claim, published whether it passes or fails.
Deviations from Capacitor Spec Rev A — flagged, not hidden
Compute: Mac mini M4 (10-core CPU/GPU, 24GB, 512GB, 10GbE) replaces the Rev A M4 Pro 48GB spec. Node aggregate drops from 432GB to 216GB unified memory. This is the cost-down interim answer to CAP-OPEN-1 (Gen 2 compute floor): sufficient for quantized mid-size models via MLX distributed inference, revisited before Phase 2 procurement.
Scale: one fluxNode, not the five-node minimum Capacitor. No external claim of "mesh" or "conformance" — the external claim is "first live fluxNode with published test results."
Masts: pilot uses 6–10m ballasted masts rather than the 10.7m permanent monopole — faster, cheaper, and typically below permanent-structure permitting thresholds (verify with Strathcona Regional District in Sprint 0).
2. Budget — $500,000 CAD
Model: leads paid, teams hybrid. People 36% · hardware and pods 46% · site, regulatory, community 13% · contingency 5%.
Line
CAD
Notes
4 contract project leads (fractional, 3 months)
$120,000
$30K each — AI, RF, Civic, Democracy II leads; includes CTO-equivalent technical sign-off for filings
Development team stipends (4 teams)
$60,000
$15K pool per team — 3–5 person cells; stipends + Class C options + DOJO/dataset access
fluxNode hardware — 3 fluxPoles
$77,000
Full BOM in Section 3; includes 9 Mac mini M4 units
Node installation and commissioning
$18,000
Rigging, alignment, electrical, grounding, testing
3 fluxPods (2 living/work + 1 storage)
$135,000
Full BOM in Section 4; NIC trades partnership build
Site — tenure, civil, pads, services
$35,000
Gravel pads, loop access, trenching, water/waste (composting + tank for pilot)
Regulatory and legal
$20,000
CBCA filing, NCLR pre-inquiry, insurance, contracts
Nations engagement and community
$12,000
Honoraria per the Early Engagement commitments; Town Hall screening costs
Contingency (4.6%)
$23,000
Thin — storage pod fit-out and one living/work pod interior are the designated flex lines
Total
$500,000
3. Hardware BOM — fluxPole (× 3)
Per-pole electronics and power, pilot configuration. Compute pricing verified against Apple Canada July 2026: Mac mini M4 24GB/512GB base $1,299 + 10GbE option $130 = $1,429/unit. All other lines are budget estimates to confirm at Sprint 0 purchase orders.
Item
CAD
Spec / purpose
3 × Mac mini M4 — 24GB / 512GB / 10GbE
$4,287
10-core CPU, 10-core GPU, 16-core Neural Engine; the AI tile compute
NEMA 4X enclosure, thermal management, tamper-evidence
$2,500
Cabinet at 3m AGL; trust-boundary requirement (CAP series)
10GbE switch + PoE injectors
$800
Intra-pole backbone for the three minis
2 × 5GHz PtP mesh radios
$1,200
Pole-to-pole full-mesh triangle links
5G cellular backhaul router
$800
WAN uplink; failover path
SDR sensing package + antennas
$1,500
KrakenSDR-class coherent receiver; RF team spectrum-as-training-signal baseline
1.8kW bifacial solar array + racking
$3,500
Per Rev A pole power spec
10.24kWh LiFePO4 + MPPT + 48V DC bus + inverter/charger
$6,000
~20h compute autonomy; DC bus scales to pod microgrid later
Ballasted mast, 6–10m
$3,500
Pilot substitute for the 10.7m monopole; no concrete foundation
Cabling, grounding, surge, connectors, misc
$1,500
Lightning/surge protection is not optional on a coastal mast
Per fluxPole
$25,587
× 3 poles ≈ $76,800 → budgeted $77,000
fluxNode aggregate (3 poles)
9 × Mac mini M4 · 216GB unified memory · 4.5TB NVMe · 90 CPU cores · 90 GPU cores · 144 Neural Engine cores · 2-of-3 pole quorum · 6+3 erasure coding target. Runs quantized models to ~30B class per node via MLX distributed inference; federated learning local-only per tile architecture. Compact cluster geometry (campground scale) — the civilian Option A/B configuration with no defence baseline geometry, keeping this pilot clean of dual-use framing.
4. Pod BOM — 3 fluxPods
One storage pod, two living/work pods, campground-loop layout: gravel pads, loop access, forest retained between, pole and battery bank adjacent to the storage pod (which doubles as the electrical/hardware room). Structural kits sourced per the theFlux Pods reference brief — BC-fibre panelized/CLT (Kalesnikoff-class supply; Nak’azdli model as precedent), assembled with North Island College trades students: the "community designed and built" clause made operational.
Living/Work Pod (× 2) — ~300 sq ft each
Item
CAD
Notes
Panelized/CLT structural kit (BC fibre)
$26,000
Factory-cut, days-not-months site assembly
Envelope — windows, doors, roofing, cedar rainscreen
$9,000
Cedar cladding ties to artboxes.ca line
Interior fit-out — murphy bed/desk, kitchenette, wet bath
$9,500
Live/work dual register: dev residency + demo unit
Mechanical — mini-split heat pump, HRV, induction, hot water
$6,000
High-efficiency, all-electric; 3–5 kWh/day design point
Electrical — 48V DC + 120V AC hybrid, LED, panel
$2,500
Shares the pole DC bus; pod battery-backup ready
Foundation — gravel pad + screw piles
$2,000
Reversible; no concrete
Per living/work pod
$55,000
× 2 = $110,000
Storage Pod (× 1)
Item
CAD
Notes
Insulated panel shell
$14,000
Single window; secure door
Racking, workbench, secure hardware cage
$3,500
Spares, tools, staging bench for pole electronics
Battery/electrical room fit — ventilation, fire separation
$4,000
Houses shared storage expansion; code-driven line
Electrical
$2,000
Foundation — gravel pad + screw piles
$1,500
Storage pod
$25,000
Pod total: $135,000
5. Organization — 4 Contract Leads, 4 Development Teams
Each lead is a paid fractional contractor ($30K / 3 months) owning one team and one gate line. Teams are 3–5 person stipend cells (open-source contributors, NIC students, advisory board). This satisfies the Rev B headline recommendation — fractional technical and policy sign-off retained in Sprint 0, not after a 4–6 month search.
Team
Mandate
3-month deliverables
AI
Tile architecture on Apple Silicon: MLX distributed inference, MoE routing across 9 minis, federated learning (local-only training), attestation prototype
Bench mesh live (Sprint 1); on-site inference across 3 poles (Sprint 4); federated learning demo; CAMA attestation handshake v0 per FIG.04
RF
5GHz PtP mesh, SDR passive sensing baseline, spectrum-as-training-signal dataset v0, link budgets, NCLR technical annex
Verified 3-pole mesh links (Sprint 3); pole-loss re-route drill; 30-day spectrum observation dataset; ISED NCLR pre-inquiry filed
Civic
theCommon registration chain (Capacitor → node → slate), fluxSlate thin-client demo, public Wi-Fi layer, community org onboarding
Registration chain v0 with public/private layer split; free public wireless live at site (Sprint 4); 3+ community orgs onboarded; Town Hall screening delivered
Democracy Level II
Deliberation design (Vervaeke-lab/SSHRC PEG alignment), CAMA governance v0.1, proportional allocation model, instrumented deliberation
CAMA governance note v0.1; deliberation protocol + metrics pack (bridging scores, decision latency, trust deltas); one instrumented community deliberation run and published (Sprint 5)
6. Three-Month Roadmap — Six Sprints, Three Gates
Sprint 0 (Weeks 1–2) — Mobilize
Sign all four contract leads. File CBCA incorporation. Adopt the canonical Phase 1/2/3 numbering across every document before anything ships externally.
Day 1: place every hardware order — 9 Mac minis, radios, SDR, solar, batteries, masts — lead time is the schedule risk, not cost. Order the pod structural kits with the NIC partnership confirmed.
Site tenure letter of intent (municipal / Hydro / Crown / private shortlist per the Lexicon site resolution); SRD permitting check on ballasted masts and pods.
Deliver the Nations early-engagement letters (Wei Wai Kum, We Wai Kai, Tla’amin) through the channels already drafted; log honoraria budget.
Publish the shared regulatory calendar (DRDC/IDEaS, CRTC status confirmation, ISED NCLR, SSHRC PEG — Sept 14 deadline lands inside this PI).
Sprint 1 (Weeks 3–4) — Bench Mesh
AI: all 9 minis on the 10GbE bench mesh indoors; MLX distributed inference demo; baseline benchmarks published internally.
RF: SDR bench bring-up; link-budget model for the site geometry; antenna plan.
Pods: fabrication starts at partner shop / NIC; site civil quotes locked.
Democracy II: deliberation protocol draft; metrics pack committed in advance (published whether favourable or not).
Sprint 2 (Weeks 5–6) — Ground Works
Site: pads, loop access, trenching, grounding grid. Solar and battery ground assembly and pre-commissioning.
Civic: theCommon registration chain alpha; fluxSlate thin-client demo on bench mesh.
Regulatory: ISED NCLR pre-inquiry filed; CRTC filing status confirmed in writing.
Sprint 3 (Weeks 7–8) — Poles Up · GATE 1
Erect three masts; align PtP radios; commission solar/battery per pole; storage pod placed and fitted as electrical room.
Gate 1: three-pole mesh links verified against link budget · node runs on solar/battery · storage pod live. Fail = Sprint 4 slips, scope does not grow.
Sprint 4 (Weeks 9–10) — Node Live
Migrate compute from bench to poles; distributed inference across the physical triangle; federated learning across poles.
Public Wi-Fi layer live (free public wireless); private-layer demo with registered slates only.
Living/work pods placed and fitted; first dev residency night — the slate comes with the front door.
Instrumented deliberation dry run with a friendly audience.
Sprint 5 (Weeks 11–12) — Prove It · GATES 2 & 3
Gate 2 (technical): 72-hour off-grid autonomy test · pole-loss drill with 2-of-3 quorum and no service interruption · results published pass or fail.
Gate 3 (civic): one instrumented community deliberation run with the Democracy II metrics pack · Town Hall trilogy public screening (phase-numbering fix confirmed first) · Phase 1 report published.
Close-out: Phase 2 funding submissions armed with published results (DRDC/IDEaS, CIRA Q1-2027 target, ICET Capital & Innovation pre-application).
7. Risk Register
Risk
Mitigation
Hardware lead times (Apple, batteries, radios)
Every order placed Sprint 0 Day 1; bench mesh work proceeds indoors regardless of site status
Permitting friction (masts, pods)
Ballasted temporary masts + screw-pile pods keep the build reversible and typically below permanent-structure thresholds; SRD confirmation is a Sprint 0 action, not an assumption
Weather / civil delays
No concrete on the critical path; pods fabricated off-site in parallel
Budget overrun (4.6% contingency is thin)
Designated flex lines: storage pod fit-out, second pod interior finish, SDR package. Compute, radios, power, and the first pod are protected
FPIC / Nations relations
Pilot site deliberately on non-reserve tenure; engagement letters delivered Sprint 0 with honoraria; no siting decision on sensitive territory without the conversation the letters open; script language ≠ engagement (Rev B flag)
Team retention on stipends
Class C options + DOJO/dataset access + residency in the pods; leads are paid — the retention risk is the cells, so keep them small and named
Overclaiming
External language is "first live fluxNode, results published" — never "mesh," "Capacitor," or "conformance." §12 Definition of Done discipline holds
8. Phase 1 Exit Criteria
One fluxNode (3 poles, 9 minis) live on solar/battery at a tenured site, with 72-hour autonomy and pole-loss quorum drills passed and published.
Public wireless layer serving the site; private layer demonstrated with registered fluxSlates; registration chain v0 operating.
Federated learning demonstrated across poles; CAMA attestation handshake v0 running per the mesh-stack figure set.
One instrumented community deliberation completed with committed metrics published.
CBCA filed · Nations engagement opened through proper channels · NCLR pre-inquiry filed · SSHRC PEG submitted (Sept 14) · shared regulatory calendar in force.
Phase 1 report published pass-or-fail — the credibility asset that opens Phase 2 (5-node Capacitor build, DRDC scale funding).
9. Related Corollaries
The Instrument gets hands — this is the first plan where the Capacitor stops being an argument and becomes a dated purchase order. The pathfinder’s return stage, rendered as a Gantt line: worth a Season Two essay when Gate 1 passes.
One node as holon — Wilber’s part/whole grammar, literalized: this single node is whole in itself (its own gates, its own demo) and simultaneously part of the five-node Capacitor it does not yet belong to. The exit criteria are the membrane.
Democracy Level II meets the dirt — Vervaeke’s relevance-realization has lived in theTotalPane’s register; Sprint 5’s instrumented deliberation is its first field measurement. The metrics pack (bridging, latency, trust deltas) is relevance-realization with a clipboard.
216 gigabytes of commons — suggested corollary essay: the node’s aggregate memory as a civic commons figure — Sheldrake’s distributed memory and McKenna’s novelty framing pointed at a rack of Mac minis in a cedar box. Closes the technical/metaphysical loop the Glossary convention calls for.
theFlux Corp · theflux.ca · icarusflyby.ca · Confidential working draft — July 18, 2026