For researchers

Papers, methodology, contestation as first-class contribution. The corpus is its own evidence — fork it, challenge it, build on it.

This page is for researchers, technical partners, open-source contributors, and serious reviewers who want to enter the research layer of FractaVolta.

For concrete deployment opportunities, see For deployers. For local Corsica-based field projects, the French operational entry point is FractaVolta en Corse.

The corpus

The work spans a public corpus of repositories, each producing research-grade documents that cross-reference each other:

• MareNostrum — strategic framework, CXU specification, DHITL (Democratic Humans in the Loop), Mediterranean solar commons, circuit-to-packet transition theory.
• FractaVolta — physical infrastructure layer: EPN, PGN, DC-native nodes, IPN, Mariani Village.
• Cogentia — cognitive infrastructure tooling: cogentia.js, Cogentia Commons methodology, Pipeline method, continuation protocol.
• inseme — platform and COP runtime: Cognitive Orchestration Protocol, briques, applications.
• barons-Mariani — political and institutional framework: Plan 2038, Discours de la seconde méthode, Projet #1755.
• Inox — language and runtime substrate: concatenative stack VM with strict control/data plane separation, designed for nodes of the future Fractanet.
• Ubikia — editorial derivation and publication layer: source-first derived products, personas, platform packages, and publication ledger.

Each repository carries a research/index.md and a research/corpus-status.md when applicable. Together they form the navigable map of the research corpus.

Foundational papers by layer

Energy substrate

• The Unconscious Grid — store-and-forward as the repressed solution to energy sovereignty; §8 extends the diagnosis to cognition, sociability, money, and territory, making this the entry point to the wider packetization corpus. A French version exists: Le Réseau Inconscient.
• Generalized Packet Networks — the cross-domain framework: heterogeneous packets, resource occupancy, congestion elasticity, packet decay, cache hierarchies, backbone/last-mile decomposition, mesh resilience.
• DC-Native Energy Packet Networks — 48 V SELV node architecture.
• Packetized Gravity Networks — hydraulic exergy.
• Thermal Packet Networks — multi-scale store-and-forward architecture for low-exergy heat and cold distribution.
• Packet Paper Template — methodological template for deriving substrate-specific papers from the Generalized Packet Networks framework.
• The Packet Transition: A Lateral Reading — phronèsis, amathia, transition taxonomy.
• The Packet as Evolutionary Attractor — complexity-theoretic framing.

Compute substrate

• CXU Specification — Compute eXergy Unit pricing primitive.
• Inference Packet Networks — packet logic applied to inference continuity.

Cognition layer

• Agent-Resumable CLI — externalized judgment, continuations, provider-neutral resumption.
• Cognitive Packets — envelope-and-payload format generalising the continuation pattern beyond the CLI.
• Cognitive Packet Switching — a protocol layer for routable ideas, continuations, and agent orchestration.
• COP Reactive Cognitive Extension — protocol surface for packet attractors, CogQueries, and pressure strategies.
• COP Implementation Profiles — documentation convention for concrete COP implementations.
• Capability Regimes — graded operating regimes for Fractanet decisions under uncertainty, urgency, mandate, traceability and recovery constraints.
• Pipeline — operational method note: pipeline on the surface, packet network in depth.
• Derived Products — source ↔ derived split: papers, blogposts, public dashboards, and institutional notes are derived products from a versioned source corpus.
• cogentia.js — Tutorial and Near-Specification — CLI concepts, storage model, workflows, and command reference.

Language and runtime substrate

• Inox — Programming Language Specification — concatenative stack VM with strict control/data plane separation.
• Reactive Sets in Inox — Native Implementation Path — implementation path from Toubkal-style dataflow toward the COP Reactive Cognitive Extension.
• The COP Manifesto — TCP/IP for intelligences.
• COP Architecture — protocol specification.
• Cogentia Commons Working Paper — methodology for distributed knowledge production.

Civic layer

• DHITL — Democratic Humans in the Loop — the axiom: AI may participate in epistemic production; living persons alone govern.
• Democratic AI Safety — the political layer of AI safety.
• Discours de la seconde méthode — methodology rules: publish process, objections as first-class, machine-readable structure, corpus as evidence.
• La méthode des terrains féconds — pre-deliberative protocol for reconfiguring oppositions before argumentative confrontation.
• Traçabilité des actes — governance doctrine of engaging acts: trace, imputation, justification, correction, revocation.
• Test du critère Rossignol — criterion for procedural stabilizers and points of external attestation.

The methodology

The corpus follows the second method:

1. Publish the process, not only the result. Continuations make the process inspectable.
2. Every objection is a first-class contribution. Falsifiable claims, marked-unfalsifiable feelings of certainty, and requests for specifics are all part of the record.
3. Structure for machine readability from the start. Stable identifiers, schemas, canonical URLs, and append-only audit logs.
4. Let the corpus be its own evidence. The repositories demonstrate the method on themselves.
5. Encode the boundary in the architecture. Epistemic participation and political governance must remain distinct.

The operational counterpart of the second method is the Pipeline: intuitions become cognitive packets; packets become versioned source documents; source documents derive into audience-specific products; public feedback reintegrates into the corpus.

The companion paper Derived Products formalises the source ↔ derived split.

The continuation protocol

cogentia.continuation.v1 is the operational realization of cop/continuation, COP's foundational continuation primitive. It is a typed, validated, provider-neutral resumption point for CLI tools.

The soundness test is binding:

Can Claude be replaced by a human, or by another agent, without modifying cogentia.js? If yes, the protocol is sound. If no, the protocol is contaminated.

The protocol is implemented in scripts/cogentia.js. See Agent-Resumable CLI for the full design.

How to contribute

The corpus is open source under CC BY-SA 4.0 for research and MIT for code.

As a contestant

• File an objection on any paper as a GitHub issue or pull request.
• Per Rule 2, your objection becomes part of the record.
• The objection can be converted into a falsifiable claim, marked as a feeling of certainty with attribution, or returned for the specifics that would make it testable.

As a contributor

• Fork a repository.
• Add to the relevant research index following canonical patterns.
• Use front matter with canonical_url when appropriate.
• Run the corpus checks where available.
• Open a pull request.

As an extender

• Write a new brique for inseme.
• Write a new continuation profile on top of cop/continuation.
• Replicate the EPN / PGN / IPN architecture on a different territory.

Toolchain

• cogentia.js — Cogentia Commons CLI: multi-repo document navigation, generated corpus views, trails, continuations, read-only GitHub issue packets, git drift and consolidation checks.
• inseme + COP — deployable platform and cognitive orchestration protocol runtime.
• .cogentia.json — per-corpus registry of repositories.
• .cogentiaignore — per-repo ignore patterns for operational artifacts that should not be flagged as unanchored claims.

Contact and next steps

• jhr@baronsmariani.org
• LinkedIn — FractaVolta

For concrete deployment opportunities, see For deployers. For local Corsica-based field projects, see FractaVolta en Corse.

All papers: CC BY-SA 4.0. All software: MIT.