BR
The Bedrock Project
Constitutional infrastructure for critical flows

Constitutional control plane · draft 0.9

A control plane for AI's physical reality — without a surveillance state.

Bedrock is a reference design for governing critical flows: energy, compute, and constrained materials. It pairs verifiable receipts with privacy-preserving proofs andfail-closed policy gates — so institutions can distinguish strategic work from waste.

Explore use casesRead the patternPrivacy & safeguards
Bedrock in one page
Rails online
The 5-layer stack
Layer 5 · InstitutionalRules, debt, capacity
Layer 4 · SocialMedia, truth, coordination
Layer 3 · HumanSkills, attention, incentives
Layer 2 · BiologyHealth, fertility, safety
Layer 1 · PhysicsEnergy, materials, grid
The core primitives
Observe

Measure critical flows with verifiable receipts.

Prove

Show rights & limits with minimal disclosure.

Constrain

No valid proof → fail closed at commit-time.

Scope: critical flows
Principle: privacy by design

From Extraction to Attribution

AI is becoming a physical industry. A world-class orchestration layer starts by making energy → compute → outcome visible — then governing it with minimal disclosure.

A short story

Once upon a time, digital systems could consume resources without a meter.

Every day, the grid grew fragile, the internet filled with slop, and scarce materials disappeared into disposable tech.

One day, it became clear the bottleneck wasn't models. It was atoms and accountability.

Because of that, we began designing measurement rails and policy gates for critical flows.

Until finally, we can move toward an Attribution Economy — where strategic compute wins over waste.

The basics
Well-to-Inference: measure the path from energy → compute → outcome so supply and demand can be governed.
Policy gates: high-stakes actions pause until required proofs exist (approve/deny/route with accountability).
Portable proofs: compact evidence packets that can be audited without exposing raw logs or private life.
Canonicals + validation: shared definitions and rules so systems agree on "what counts," at decision time.
Integrity rails: Artifact Integrity (is it real?) and Agent Integrity (is it safe to trust?) for high-stakes workflows.
Not a surveillance project

Bedrock governs critical flows using minimal disclosure and privacy-preserving proofs — not total life-logging.

The polycrisis is an attribution failure.

We talk about energy, AI, and trust as separate crises. Underneath, the root cause is the same:we cannot see who did what, with which resources, under which rules — in time to correct course.

When purpose is invisible, rationing becomes panic. When provenance is absent, slop wins. When midstream is opaque, rivals gain leverage. Bedrock is a measurement and enforcement substrate designed to restore feedback loops — without collapsing into surveillance.

Failure patterns
the compounding loop
Curtailment → Grid stagnation: we dump clean power instead of using it.
Opaque compute → Grid fragility: operators see load, not purpose.
Model slop: systems train on their own exhaust; truth and creativity blur.
Midstream abdication: we mine at home, refine abroad, and depend on rivals.
Youth betrayal: blocked housing, fragile work, and no upside in the status quo.
What's missing today
the missing layer
Meters

Verifiable receipts for energy, compute, and constrained materials — at decision time.

Proof

Privacy-preserving proofs of rights, duties, caps, and provenance.

Gates

Automatic, fail-closed enforcement when required evidence is missing.

Integrity

Artifact Integrity (origin) and Agent Integrity (behavior) for high-stakes workflows.

Use case library

Concrete, repeatable patterns. Each one is the same shape: receipts → proofs → gates → portable evidence.

Use cases below are illustrative starting points—not an exhaustive catalog. The Bedrock layer applies to any workflow that touches critical flows (energy, compute, materials, integrity, rights).

6
Core patterns
4
Integrity rails

Grid draw governance

Prioritize hospitals and critical services when power is constrained — based on purpose, not hype.

Well-to-Inference
kWh receiptspriority proofsfail-closed caps

Training provenance

Require provenance and composition evidence before a model can train at scale.

Slop control
dataset receiptshuman/synth labelsproof packet

Material passports

Prove allied midstream and lawful refinement for critical programs.

Supply chain
origin receiptsrefining proofscontract gates

Data & resource royalties

Attribute value so citizens can share upside when resources are consumed for intelligence.

Attribution
usage receiptssettlement rulesaudit proofs

Artifact Integrity

Verify origin and tamper-evidence for content in high-stakes workflows.

Integrity
origin proofschain-of-custodydistribution gates

Agent Integrity

Fail closed when systems attempt manipulation, coercion, or policy-evasion.

Integrity
behavior gatespolicy versionsoperator resolution

The Bedrock Layer: receipts, not revolutions.

Bedrock is a constitutional control plane that sits under critical infrastructure. It doesn't replace markets or votes — it makes high-stakes flows observable and enforceable where it matters. The primitive is always the same: a verifiable receipt attached to a unit of flow, then evaluated by policy at commit-time.

01 · Observe

You cannot govern what you cannot see.

Instrument critical flows with verifiable receipts: where power was generated, how compute was spent, where materials were refined.

02 · Prove

Prove rights and limits, reveal nothing extra.

Demonstrate compliance using minimal disclosure: permitted, within caps, using allowed sources — without exposing raw logs or private life.

03 · Constrain

No valid proof → no run.

If required receipts or proofs are missing, the system fails closed and routes decisions to accountable operators.

Governance alignment, made enforceable.

Most frameworks describe what "good" looks like. Bedrock turns those principles into execution:receipts for evidence,proofs for verification, andfail-closed gates for enforcement.

Designed to support governance programs aligned with NIST AI RMF, ISO/IEC 42001, OECD AI Principles, and UNESCO Ethics. (Not a certification claim.)

Framework compatibility
  • NIST AI RMF: Govern / Map / Measure / Manage become concrete control points inside workflows.
  • ISO/IEC 42001: an evidence layer for AI management systems (policies, monitoring, improvement).
  • OECD: traceability + accountability via receipts and decision records; safe override/stop via gates.
  • UNESCO: supports monitoring, transparency, and responsibility through provable lineage and intelligible decision context.
  • EU transparency duties: helps label inauthentic content and preserve "why" for high-stakes decisions.
AI safety primitives

Safety is a physics problem when AI touches critical flows. Bedrock adds integrity rails that fail closed when the system cannot prove trust.

Artifact Integrity
Prove origin and detect tamper / inauthentic media where trust matters.
Agent Integrity
Detect policy-breaking behavior (manipulation, deception, unsafe action) and route to accountable human resolution.

The goal is not "more policy." The goal is fewer ways to cheat the policy.

Genesis Mission alignment

Genesis Mission is about accelerating discovery with secure, interoperable compute + data. Bedrock strengthens that stack by making critical flows measurable and governable across partners.

  • Accelerate discovery with reproducible, auditable lineage for data, models, and runs.
  • Strengthen national security via fail-closed access, vetting hooks, and tamper-evident audit trails.
  • Secure energy dominance through Well-to-Inference measurement (energy → compute → outcome).
  • Enable partnerships with portable proofs that satisfy security, IP, and data-use constraints.

Privacy is not an afterthought. It is the architecture.

The obvious fear is that attribution rails become mass behavioral scoring. Bedrock is designed to prevent that outcome through scope limits and minimal disclosure.

Critical flows, not total life-logging.
We care about systemic risk surfaces (grids, supply chains, high-stakes systems), not personal communications.

Proofs, not dossiers.
Actors prove compliance (caps, rights, provenance) without exposing underlying identities or full histories.

Constitutional constraints by design.
The rails that enable governance should also restrict repurposing for mass scoring.

Cheap metaphor

A bartender doesn't need your full identity to serve you. They need a single bit: over / under 21.

Modern systems demand the whole ID. Bedrock demands only the bit — proven via cryptography, not trust.

What institutions see
  • Risk posture and compliance at the system level.
  • Which critical thresholds are being approached or breached.
  • Where sovereignty is threatened (e.g., midstream chokepoints).

Not: individual browsing histories, DMs, or private conversations.