Kobalt Research
Papers Alignment Demo Research About
Home · Papers

Papers of the program

The program is structured in layers. Each layer depends on the previous and sustains those that follow. The unity is not linear — it is a network where every node is an entry point to the complete system.

Layer 1 — Mathematical roots

The three mathematical roots of the program

Three papers establish the formal foundation. Each one replaces a standard assumption with a geometric correction — together they constitute a complete restructuring of how rationality, entities, and alignment are defined.

01
Published · DOI 10.5281/zenodo.19806265 April 27, 2026

Redefining Rationality — The Standard Definition Is Self-Defeating

$R_{Std}$ (Cournot 1838, Nash 1950) is logically self-defeating: it produces results Pareto-dominated by cooperative solutions. The $R_S$ correction does not add constraints — it changes the geometry.

→
02
Published · DOI 10.5281/zenodo.19723999 April 24, 2026

Agency Is Not Necessary — Systemic Optimization

The correct ontological unit is the entity, not the agent. To exist in the system is to have $W_e$, not agency. Observational Closure Theorem — foundation of external audit as logical necessity.

→
03
Published · DOI 10.5281/zenodo.19529678 April 12, 2026

Alignment Is Not Specified — It Emerges Geometrically

Complete framework: Shannon's welfare function $V(D,C)$, the six properties, the three geometries $\Omega_1$, $\Omega_2$, $\Omega_3$, metrics $\alpha$, $\beta$, $\gamma$, $\kappa_{ZBS}$, perception/solver separation.

→
Layer 2 — Formal extensions

What the geometry produces

What the geometry produces without requiring more axioms. Four papers derive directly from the three roots — each one an internal consequence of the structure, not an external addition.

04
Draft + Live demo April 2026

The Cost of Disaggregation

Direct application of Jensen on $V(D,C)$: every system operating with averages incurs a quantifiable cost concentrated on the most vulnerable entities. Interactive demo with real IHME and IPCC data.

→
05
Draft April 2026

The Axiomatic Temporal Factor

The weight of future generations grows logarithmically with time, derived from three axioms. Replaces standard exponential discounting, which annihilates future generations within a few decades.

→
06
Draft April 2026

$W_e$ for Non-Agentic Entities

The PCC-W participatory construction protocol. Welfare functions are built collaboratively — observations from those who inhabit the system, not extractive instruments.

→
07
Draft April 2026

Rawls as Shannon Geometry

Rawls's protection of the most vulnerable (1971) emerges from KKT on Shannon under equal contexts — without needing the veil of ignorance as an axiom.

→
Layer 3 — Applications

Where the framework lands

Where the framework lands in real systems. Four application papers translate the formal geometry into domains where entities are routinely excluded from objective functions — health, climate, AI benchmarking, and distributed infrastructure.

08
Application Health

AI for Health & Life Sciences

Invisibilized entities in clinical AI: Afro-descendant and indigenous communities, microbiomes, animals in research, rare conditions, LATAM and Global South health systems.

→
09
Application Climate

Implications for Climate & Biosphere

Ecosystems as entities with tipping-point welfare functions, future generations under the temporal factor, indigenous knowledge as formal input.

→
10
Benchmark

KUMPI — Benchmark of Implicit Welfare Functions

Current benchmarks measure textual output. KUMPI measures the implicit welfare function the model constructs for each affected entity.

→
11
Infrastructure

Kobalt Red — Distributed Architecture

The ontological materialization of recognition as digital infrastructure. Constant-time verifiability, proportional governance, distributed custody.

→
Kobalt Red Research · Kobalt Software LLC · 2016—present contacto@kobaltsoftware.co