The commercial proposition

Thirty years of asking "how do autonomous agents coordinate without a capturable centre?" — in programming languages, sensor networks, energy infrastructure, and democratic institutions. We know how to answer it as a stack. We charge for the engineering, integration, and operations — not for the answer.

We do not sell electrons on a grid, carbon credits, generic consulting, or closed-source SaaS lock-in. We do not license the software (it's MIT). We sell the deployment, the integration, the operations, and the regulatory and engineering judgment that makes the stack run on a real site.


What you buy

Energy substrate

  • Feasibility studies for site-specific Energy Packet Network deployments — PV sizing, battery dimensioning, packet flow analysis, regulatory navigation (IMDG maritime, SELV electrical codes, RGPD for any installed sensing).
  • DC-native node design — 48V SELV bus, MPPT, LFP, edge compute, USB-C PD 3.1 interfaces.
  • Battery container engineering — 2 MWh ISO 20' LFP packets, certified to maritime freight standards, BLE-tracked, ferry-routable.
  • PGN integration where applicable — historical irrigation canals as gravitational storage substrate, IEV (Inverted Energy Vessel) node design, Corsica case study materials available.

Compute substrate

  • CXU-priced sovereign inference — Compute eXergy Unit accounting: energy × hardware efficiency × system efficiency × SLA premium = one auditable price per useful inference. See CXU_SPEC.
  • Solar-backed edge inference nodes — design, deployment, operations. Stranded Mediterranean exergy → inference tokens at ×10–×40 the revenue per MWh of grid export.
  • Inference Packet Networks (IPN) — bounded continuity layer for when hyperscale AI is degraded, unavailable, geopolitically disrupted, or energetically constrained. See the IPN paper.

Integration

  • Stack assembly — deploying the cognition layer (cogentia.js + Cogentia Commons + continuation protocol) and the civic layer (inseme) on FractaVolta hardware.
  • Multi-instance configuration for per-commune / per-site deployments of Kudocracy.Survey and Inseme Agora.
  • Ophélia AI mediator integration with sovereign inference backends — no vendor lock.
  • Audit and observability — every layer produces typed events; the audit trail is a first-class deliverable, not a compliance afterthought.

Operations

  • Operational training for site staff
  • Deployment support
  • Cogentia Commons editor role (if you want the corpus methodology applied to your domain)

Value capture model

The pricing logic is avoided interruption at every layer:

  • Energy: avoided downtime when the SARCO cable saturates or the grid trips. We sell the continuity guarantee, not the energy.
  • Compute: avoided inference outage when hyperscale degrades. The Inference Packet Network paper makes this explicit — premium cognitive backup priced by avoided cognitive interruption, not nominal token throughput.
  • Cognition: avoided reasoning loss when an AI provider changes terms, retires a model, or goes offline. Continuations preserve and replay the work.
  • Civic: avoided platform capture when a SaaS vendor changes terms or a political wind shifts. inseme is yours forever — it's MIT, it runs on your hardware, it produces auditable events.

This is a service model, not a licensing model. The product is integration and operations.

What you also receive — the audit trail

Beyond the hardware and the running platform, every project ships with its own audit chain — produced live during the engagement, not assembled retroactively.

  • Decisions are typed, via the cogentia.continuation.v1 protocol implemented in cogentia.js. When a deployment decision required judgment (which battery topology for site X; which inverter for that grid-tie context; how to size the inference cluster), the question, the alternatives we considered, the chosen path, and the reasoning are all preserved as JSON in the project's .cogentia/continuations/. You can replay them.
  • Failed alternatives are not erased. When we tried something and it didn't work, the failed branch is part of the record — with the reason. The next operator (us in two years, or a different firm if you change supplier) can read why we ruled options out.
  • Claims are anchored. Every datasheet number, every architectural assertion, every cost estimate points at a canonical document under version control. If we change a number, the change is visible in the git history.
  • The AI we use is replaceable. Any continuation we emit can be resumed by another agent — Claude today, ChatGPT tomorrow, a local model next year, a human operator at any point. No provider lock-in flows down to you through us.

This is not compliance theater layered on top of normal work. It is how the work is done. See the methodology page for the formal account.


Pilot opportunities

  • Mariani Village pilot — Tavignano valley, Corte. ISO container DC-native housing fleet, seasonal routing Corte ↔ coast, full stack deployment (energy + compute + cognition + civic) on a single site. Operated by Dilorta SAS.
  • PGN pilot — Tavignano valley hydraulic infrastructure, IEV node feasibility, gravity-as-territorial-memory case study.
  • Sovereign inference cluster — solar-backed Corsica node, IPN demonstrator, CXU-priced inference services for regional partners.

See partner_brief.md for the detailed engagement brief.


How to engage

We work with:

  • Public-sector partners seeking sovereign infrastructure (collectivités, regional authorities, public universities)
  • Cooperative housing and community land trusts deploying off-grid residential infrastructure
  • Industrial sites with high-value local compute needs and stranded renewable generation
  • Academic institutions running deliberation infrastructure for student governance or research consultations

We do not work with operations seeking vendor lock-in, opaque audit chains, or AI-as-extraction monetization.


Contact

📧 jhr@baronsmariani.org 🔗 LinkedIn 🌐 FractaVolta on GitHub

Engineering: Institut Mariani. Commercial vehicle: Dilorta SAS (housing). Knowledge governance: Cogentia Commons.