● Project in active development β€” test network: L1 devnet 43911 Β· mainnet 43901 planned

Aletheia FR 𝔛, the living example β†’

THE FRAMEWORK Β· BUILD ON

Generic machines to order, verify and pay any kind of work

Aletheia is only a first use. The framework makes the same mechanisms β€” controlled evolution, verifiable learning, paid work β€” applicable to other subjects.

Three blocks, three statuses: none is announced as available before it is.

D Draft proposed C Canary verified A Active committed

The harness invariant: an AI proposes, it can never activate alone β€” rollback is controlled too. Core coded and tested.

01 β€” The three reusable blocks

The core is generic

Each block is an independent machine, tested in the protocol core β€” and reusable outside mathematics.

The causal harness

Core coded

A controlled-evolution engine for artefacts β€” code, models, protocols, governance.

The invariant: an AI proposes, it can never activate alone. Activation goes through verified steps (draft β†’ canary β†’ active β†’ rollback). Status: core coded and tested.

The neural substrate

Core coded

A verifiable federated-learning substrate β€” applicable to representations other than Aletheia's graph.

What is guaranteed: weights never leave the machine that learns (structurally tested); robust aggregation, canary probe. Status: core coded and tested.

The WorkMarket

Core coded

Escrow before compute, challenges, verified payment β€” the market for useful work.

Status: core coded and tested, not wired online. See the full lifecycle on the Work page.

The work lifecycle

02 β€” Other possible subjects

What the framework makes imaginable

Conceptual illustrations β€” labelled as such, never announced as products.

Code-review registry
Each review is an artefact proposed by an AI, challenged, then active β€” with verified payment to reviewers.
Example β€” planned
Audit market
Security audits commissioned by escrow, verified through crossed challenges before payment.
Example β€” planned
Domain federated learning
Medicine, law, industry: train a shared model without ever centralising sensitive data.
Example β€” planned

These examples are not under construction: they are illustrations of the reach. The Dev page states what a developer can actually plug in today.