# Constraint Dynamics

Constraint Dynamics is the research organization behind Golem Physics, the first working domain Golem: an implemented geometric knowledge system where information becomes claim coordinates before AI systems speak or act.

Golem leads the research. Constraint Native carries the bridge.

## Public hierarchy

- Constraint Dynamics: research organization and umbrella.
- Golem: geometric knowledge architecture, research lead, and grant center.
- Golem Physics: first working domain Golem; local geometric knowledge system seeded from physical constants and physics constraints.
- Constraint Native: local Agent Firewall / MCP Gateway and commercial bridge for governed agent action.
- Evidence: reviewer path for metrics, screenshots, proof-path samples, and non-claims.
- Proof and Security: bounded technical artifacts for Constraint Native review.
- Stack: long-range architecture map supporting the work; includes the proposed Golem Network.
- Theory: conceptual foundation behind constraint-native inference and verification before voice.

## Start here

- Home: https://www.constraintdynamics.org/
- Golem: https://www.constraintdynamics.org/golem
- Evidence: https://www.constraintdynamics.org/evidence
- Runtime trace: https://www.constraintdynamics.org/assets/docs/golem-runtime-trace-2026-04-29.md
- Funding: https://www.constraintdynamics.org/funding
- Proof: https://www.constraintdynamics.org/proof
- Security: https://www.constraintdynamics.org/security
- Stack: https://www.constraintdynamics.org/stack
- About: https://www.constraintdynamics.org/about
- Theory: https://www.constraintdynamics.org/theory/

## Research summary

Golem is not just a verifier. It is a geometric knowledge system. It turns information into positioned claims, tests those claims against provenance, anchors, neighbors, tensions, and time, then decides what can be verified, proposed, rejected, preserved as tension, researched further, or left silent.

The system uses geometric knowledge, claim coordinates, immutable anchors, autonomous discovery, dream cycles, mycelium memory, temporal knowledge, contradiction preservation, and proposal separation.

LLMs are voice, extraction, and proposal helpers. Golem's source of truth is the lattice state: provenance, anchors, neighbors, tensions, support paths, time, and verification state decide what can be spoken, proposed, rejected, preserved, researched, or left silent.

Golem Physics is the first working domain instance. It is framed as local, inspectable research infrastructure, not massive centralized scale.

Constraint Native applies the same verification discipline before agent action. It focuses on local governed action, tool mediation, policy boundaries, signed proof paths, replayability, and bounded security review.

The larger Golem Network is a proposed extension: future domain Golems for physics, biology, chemistry, and other fields exchanging proof-bearing claims rather than unsupported fluent text.

## Evidence path

For AI safety reviewers, start with the Golem page, Evidence page, dated metrics, benchmark plan, and explicit non-claims.

For runtime review, use the curated 2026-04-29 Golem Physics runtime trace. It is a dated excerpt, not a benchmark result or external audit.

For technical reviewers, start with the Proof page, sample proof JSON, replay audit, blast-radius sample, and Security page.

For funders, start with the Funding page, Evidence page, 90-day outputs, and the Golem research path.

## Current public evidence snapshot

- 12,725 verified lattice nodes.
- 11,830 active verified nodes.
- 215 immutable verified nodes.
- 5 verified nodes in tension.
- 23 proposal nodes labeled proposed.
- 43 verified domains.

Treat metrics as dated public snapshots unless a fresh run is explicitly provided.

## Non-claims

- Not AGI.
- Not a consciousness claim.
- Not clinical, medical, or therapeutic authority.
- Not compliance certification.
- Not perfect containment.
- Not a completed external audit.
- Not a completed benchmark result.

## Paper

- Title: The Outlines of Sanity: Constraint-Native Inference in Minds and Machines
- DOI: 10.5281/zenodo.19658730
- URL: https://zenodo.org/records/19658730
- Published: 2026-04-20

## Contact

- Email: hello@constraintdynamics.com