The documents behind the design.

Why blockchains did not move like AI, and a design that sells trust rather than ordering: two-sorted state, Linear Lanes and Contention Consensus, data availability, the Fold, the Assurance Algebra, Verifiable Clearing, Agent Charters, privacy and compliance by proof, the security model, the economics, six applications with their baselines, a phased plan with kill criteria, and open research problems.

The paper describes a system to be built. Its capacity and latency figures are design targets, and the reference implementation covers part of the design.

The function a proof stack evaluates for every chunk of every Pulse, and through it the whole state transition: preparing and executing a transaction, every operation, the three phases of a Pulse, the commitments a chunk statement carries, verifying a chunk and folding a Pulse. Proof stack B was written from this document alone.

Version 1.3 added validators, stake, epochs, rewards, slashing of equivocation, forced-inclusion obligations and committee rotation. Version 1.4 pays rewards from the fees collected during each epoch, and from nothing else. Version 1.5 takes their seats from slashed validators at the epoch boundary even when nobody else is eligible. Later versions add, among other things, payment streams, agent identities and subagents, memory logs, Task Charters, storage deals and one signature for a batch of keyless subagents.

What the reference implementation builds, simplifies, simulates or leaves out, section by section of the architecture, the agent on your computer included; the trust assumptions it adds; how a light client is trusted; and what the independent reviews found. Every claim on this site is held to it.

Where the architecture breaks, and what would have to be true for it to go much further: billions of payments a second, billions of agents, succinct validity proofs, networking between machines, and agents that are reliable and improve themselves. Each note says what is built, what is measured and what is only a plan.