// architecture of self-auditing cognition

Kevin Kull

Creator of M-COP 2.0 — a framework for verifiable, self-auditing AI cognition. I build production-grade systems where reasoning is traceable, energy is conserved, and the seam between mind and machine is engineered, not assumed.

New Brunswick · UTC-3 Open to research collab 32 repos · 12 followers

What I'm building

§ 02  ·  thesis

Cognition should be auditable. Energy should be conserved. Causality should be respected. Everything I work on lives at the intersection of those three constraints — and the surprising shapes that emerge when you take them seriously.

Domain
Self-auditing AI cognition
Posture
Source-available · BUSL 1.1 → MIT 2030
Currently
M-COP 2.0 · v0.4-rc · stable
Method
Verifiable, production-grade, reproducible

Pinned work

§ 03  ·  03 of 32 repos

Active lab

§ 04  ·  live signal
/ audit · stream

The cognition logs itself.

M-COP 2.0 surfaces a continuous trace of its own reasoning steps, energetic cost, and constraint violations. The terminal below is a sample window into that stream — sanitized, slowed, and stitched from real runs.

/ ledger · trajectory

How we got here.

A timeline of the work, not a CV. The questions are what continued.

2026M-COP 2.0 — self-auditing cognition, v0.4-rcframework
2025Sensory-Tracer Science — proposal for a new subfieldresearch
2025thermo-truth — interpreting the cyber-physicalprototype
2024M-COP 1.0 — recursive meta-cognitive optimizationframework
2023Verifiable inference, energy-bounded agentswriting
TypeScript Python Rust CUDA Formal methods Thermodynamics Causality
// channel · open // sig · 0x4B4B

Send a signal.

I'm reachable for research collaboration, framework integration, and unreasonably ambitious projects at the seam of physics, computation, and cognition.