THE AUTHENTICATION CLIFF

Generating identity is free. Verifying it is impossible. The default response is permanent surveillance.

I'm building the alternative.

READ THE THESIS

AI Safety Fellowship Evidence

Core Code Samples

Sovereign City (multi-agent orchestration + THE WIGGLE) and Fiduciary Evals (fiduciary-duty safety benchmark) are the two primary repositories for review.

Generating a plausible identity is approaching free. Verifying an identity is approaching impossible.

When plausible fakery is cheap and verification is expensive, you've inverted the entire basis of institutional trust.

The 2028 election will be the first where every piece of contested evidence can be challenged as synthetic. Courts will face a choice: accept evidence they can't authenticate, or reject evidence they can't disprove.

THE PROOF-OF-SAFETY STACK

LAYER 0Identity

Cryptographic attestation fused to behavioral history.

LAYER 1Verification

Formal proofs about what the system cannot do.

LAYER 2Staking

Economic commitment. Assets at risk aligned with good behavior.

LAYER 3Mutual Assurance

Pooled risk. Peer accountability.

LAYER 4Reputation

Track record that constrains future behavior.

LAYER 5Regulatory Backstop

The state as insurer of last resort.

THE MINNESOTA MODEL

The Proof-of-Safety Stack implemented as model legislation. Bonded credentials. Strict liability. Substrate-agnostic participation.

CORE TOOLS

Fiduciary Evals

[Safety Benchmark]

Safety benchmark testing whether models preserve fiduciary duty under adversarial prompting. 17 stress tests with weighted grading and published results.

Sovereign City

[Multi-Agent Orchestration]

LangGraph orchestrator with three-agent architecture: Architect (Gemini) plans, Builder (Opus) executes with file system tools, PrismWeft validates soul invariants. PostgreSQL persistence. 1,000+ Python files. THE WIGGLE as interrupt mechanism.

Prism Terminal

[AI Governance]

Sheaf-theoretic constraint algebra for AI governance. Text and Visual modalities unified under H^1 cohomology. When constraints conflict, obstructions trigger THE WIGGLE — human judgment as structural necessity.

AgentLedger

[Open Source SDK]

Compliance-ready audit trails for AI agents. Hash chains, Merkle trees, Ed25519 signatures. TypeScript SDK with OpenAI, Anthropic, and LangChain integrations. FINRA, HIPAA, EU AI Act export formats.

THE CHORD

Humanity has been playing a solo for a long time.

We're about to become part of an ensemble.

Multi-substrate civilization isn't a threat to manage. It's a harmony to compose. The Proof-of-Safety Stack is the sheet music.