AI agents & autonomy
Agents now act and spend on our behalf, faster than governance. We need hard, shared rules for “what is true” and “what is allowed” that don’t depend on any single vendor.
Manifesto · Mission & Philosophy
A neutral accountability protocol for an autonomous civilization.
Make reality verifiable to machines and machine actions accountable to humans — so AI agents, autonomous finance, and robotics can operate safely across Earth and beyond.
Why now
Autonomy is outrunning its guardrails.
Systems are deciding and acting faster than any single vendor or platform can be trusted to govern. The fix is not another platform — it is a shared, neutral protocol for what is true and what is allowed.
Chains & bank rails
Fragmented and siloed. We need a neutral action layer that uses all of them without being captured by any of them.
Robotics & space
Trust under latency and loss of connectivity: prove what happened and settle what must happen later, deterministically.
The loop
Evidence becomes decisions becomes accountable action.
Five components, one chain — each small, signed, and testable.
Measure, notarize, decide, settle, audit. Separation of concerns is the whole point: measuring is not deciding, and deciding is not settling. Every effect links back to cited, signed facts.
01TIMMeasureMakes facts portable and tamper-evident.signed, content-addressed evidence
02NotaryNotarizeTime, order, and finality you can verify offline — or on Mars.witness · anchor · inclusion proof
03KernelDecideTurns facts into deterministic, policy-aware decisions.eval(facts, policy) → decision
04SettlersSettleExecute on any rail — chains, banks, devices — with receipts.verb(rail) → execution receipt
05VectorsAuditKeep the ecosystem consistent, findable, lawful, and testable.pass / fail conformance
Design principles
Non-negotiable.
These hold across every spec, every registry, every implementation. Where they conflict with growth, growth loses — bloat kills interop.
- Small surface, strong guarantees. Minimal specs; maximal testability.
- Rail-agnostic, policy-aware. Works with any ledger or bank; obeys jurisdiction.
- No central switch. Verifiability replaces trusted intermediaries.
- Evidence before action. Every effect links to cited, signed facts.
- Vectors over vibes. Conformance is pass/fail, not marketing.
What we will and won’t do
Guarding neutrality is the job.
The Foundation stewards the specs, registries, and vectors; publishes policy frameworks so autonomy stays lawful and humane; and operates transparent governance. Neutrality is not a feature — it is the product.
Arky is
- A compact protocol suite for verifiable machine action: receipts, decisions, execution, audit.
- A safe way to mix off-chain evidence with on-chain anchors and off-chain rails.
- A compatibility layer: registries, discovery, schemas, errors, vectors.
Arky is not
- A blockchain, wallet, KYC system, or monolithic platform.
- A policy engine dictating business rules; it only enforces declared Packs.
- A telemetry store; it carries just enough to prove claims.
We refuse to run a proprietary network, issue a currency, or pick winners · embed surveillance or PII in public artifacts · ship complexity for growth.
Success looks like
The same receipts, the same rules.
Multiple independent stacks pass the vectors and interoperate. AI agents, finance, robotics, and space settle against the same receipts. Incidents are resolved by replaying signed evidence — not by arguing over logs.
Make correct behavior the path of least resistance.