FAQ

What is Siontology?

An open-source reference implementation of a neo-blockchain: a network whose product is verified knowledge. Every claim it records carries a proof or priced evidence. See Neo-blockchains.

It keeps a replicated ledger with validators and consensus, but it is organised differently. Knowledge is appended without coordination; only the consumption of resources is ordered, and only among contenders. Owned transfers finalize with one committee certificate and no consensus round. Every Pulse is proven, so a client can check validity without replaying history.

Can I use it in production?

No. It is a devnet-scale reference implementation: a devnet runs every validator in one process (separate sion node processes also talk over QUIC, tested on one machine), the proofs are replay proofs that simulate succinct ones, and devnet keys can be derived by anyone. It secures nothing of value. See What exists today.

Can I run an agent of my own with it?

Yes, on your computer: sion setup makes a chain of your own, your wallet and your agent’s, the agent under a charter, its memory log, its model and a profile, and sion talks with it. It answers in the terminal and on messaging services, works in a folder you give it, browses, remembers, and asks before it spends. It has been tested with a scripted model and stand-ins for the model providers’ and messengers’ APIs, and no release has been published yet, so the installer builds it from source. See Run it on your computer and Your agent.

Is there a token?

Siontology has no token and nothing is on offer. Fees are paid in stablecoins.

Who are the validators, and how do they join?

On a devnet, four genesis validators, plus any standby validators you start. A validator joins by staking, takes its seat at the next epoch boundary, earns a share of fees, and loses its stake if it signs two conflicting messages. See Validators, staking and epochs.

Is it post-quantum?

Every account and validator signs with two unrelated post-quantum families at once, ML-DSA-44 and SLH-DSA-SHA2-128f, so a forger must break both, and payloads are encrypted with ML-KEM-768. The implementations are unaudited pure-Rust crates (impl:unaudited-pq-crates). See Post-quantum identity.

Are the proofs real?

The World Proof’s bindings are real: it chains every chunk’s statement, binds receipts, published data and lane certificates, and two independent implementations check every chunk. The proofs themselves are replay proofs: verifying one costs as much as executing it. A succinct proof stack would plug into the same interface; none is built yet. See The Fold and World Proofs.

Why two proof stacks?

So that a bug in one cannot forge the world: a World Proof verifies only if both accept every chunk. The second stack is written from the specification alone, and their disagreements have already found real defects. See Keeping proof stack B independent.

How fast is it?

In one benchmark run on a shared 2-core machine, a devnet of four validators closed a Pulse every 3.9 seconds and moved about 15 transfers a second, mostly bound by post-quantum signatures. The architecture’s targets (100,000 operations a second on a testnet) assume many machines and signature aggregation in proofs, and have not been measured. See Benchmarks.

Can it do billions of transactions a second?

Not as general transactions: our own estimate for a billion a second with succinct proofs is millions of prover machines. It can settle payments at that rate. With payment streams, each payment is checked by its payee with one hash, and a settlement by checkpoint costs the ledger the same however many payments it covers. On one core of a 2-core sandbox, settlements covering about 9 × 10⁸ payments a second executed with every signature checked. A billion payments a second would then need about a thousand settlements a second on the ledger; that has not been run across machines. See Billions of payments a second for the measurements, the limits and what is not new.

What can a phone check?

With the chain identity pinned: that a header was certified by the committee of its epoch, that a receipt, an object or a claim is in a certified Pulse, that the Pulse’s data is available (by random sampling), and, at the cost of replay, the Pulse’s World Proof. See Verify as a light client.

How does an AI agent use it?

An agent of your own runs in the runtime that comes with it (Your agent). An agent built elsewhere uses the MCP server (sion mcp), the CLI or the HTTP API. Either can pay inside an Agent Charter whose limits the ledger enforces, price a claim before relying on it, keep a memory log, post and settle work, and buy HTTP resources through x402, one transaction per request or one hash per request from a payment stream. See Connect an AI agent over MCP. Programs and other agents reach an agent of your own over HTTP, OpenAI’s chat API and A2A (Channels).

What if my transactions are ignored?

Post them as forced intents (sion send --forced). One honest proposer is enough to order them, and a shard that does not execute them in time has its committee replaced and the miss recorded in every certified header. See Censorship resistance.

Can I run validators on several machines?

Not yet across machines. A devnet’s validators exchange real signed messages over in-memory channels inside one process (impl:in-process-network). Separate sion node processes talk over QUIC and have certified and ordered transactions together on one machine, with peers read from files and no way for a stopped validator to rejoin (net:static-peers, net:no-rejoin). See Running a network of processes.

Siontology

FAQ