Chain
Open Primitive

ZK Proof + Useful Computation.
A New Kind of Chain.

The idea is simple. Take a ZK proof system. Tie it to a computation that produces something real. Make that the heartbeat of a blockchain. Every block becomes a cryptographically signed fact about the world. Markovian is the first chain built on this primitive. A verifiable classifier is the first application. The primitive itself is open.

The Primitive

What makes this different

Bitcoin aimed SHA-256d at nothing. It produces security. It produces a ledger. But the computation itself, the hashing, is discarded. It proves work was done. That is all.

Markovian changes the target. Miners still run SHA-256d. But they also run Hidden Markov Model Viterbi computation, ZK-verified per block. The proof is part of block validity. The chain state is the output of that computation, a live verifiable classifier, updated every block, cryptographically verifiable by anyone.

The computation is no longer discarded. It is the product.

# Every Markovian block contains: block.sha256d_nonce # standard proof of work block.viterbi_output # HMM state classification block.zk_proof # BN128 Pedersen + Schnorr sigma via Fiat-Shamir block.zk_valid # on-chain verification result block.chain_state # the signed output: State 0 | State 1 | State 2
The Template

The computation graph is swappable

Markovian uses HMM Viterbi because its fixed computation graph, no branching, predictable depth, maps cleanly to a purpose-built sigma protocol. Proof generation is fast enough to run at block frequency. Verification is cheap. The design holds.

But the protocol does not require Viterbi. It requires a computation graph with a verifiable output and a ZK proof that fits inside a block. The verifiable classifier is instance one. The template is open.

Instance 01
Verifiable State Classification
HMM Viterbi over a public observation sequence. Chain state is one of three signed states, State 0, State 1, or State 2. Live since June 4 2026. 118,000+ blocks.
Possible
Weather State Classification
HMM over atmospheric sensor data. Chain state is a cryptographically signed weather regime. Immutable, auditable, no trusted data provider.
Possible
Genomic Sequence Alignment
Viterbi over DNA sequences. Every block archives a ZK-verified alignment result. A growing dataset of verifiable biological facts.
Possible
Satellite Imagery Classification
ML classifier over remote sensing data. Chain state is a verifiable land-use or climate signal, updated every block.
Possible
AI Model Inference Log
Fixed-depth inference graph ZK-verified per block. Every block is proof that a specific model produced a specific output from a specific input.
Possible
Oracle-Free Settlement Layer
Any computation whose output can settle a contract. Prediction markets, parametric insurance, derivatives. Settlement from chain state. No manipulation surface.
Why It Works

The ZK proof is what gives it life

Without the ZK proof, this is just a chain that claims to run ML computation. With it, every block carries cryptographic evidence that the computation ran correctly. The output is not trusted. It is verified.

Markovian uses BN128 Pedersen commitments with sequential Schnorr sigma proofs via Fiat-Shamir. For a fixed computation graph like Viterbi, no branching, predictable depth, this outperforms general zkVM approaches at block frequency. Verifier efficiency was the design constraint. The proof fits.

The full ZK specification is in the whitepaper. The proof construction, the commitment scheme, and the verification logic are all documented.

Honest Scope

What oracle-free settlement does and does not cover

Oracle-free settlement works when the settlement condition is derivable from chain computation. If the answer lives inside the chain's math, no oracle is needed. If it lives in the real world, you still need one.

Markovian settles this
The answer is chain state. No external report needed.
Will the chain be in State 1 by a target block?
Is the current block in State 2?
Has the chain held State 0 for 10+ blocks?
State-triggered settlement contracts
Parametric payouts keyed to chain state
Divergence between two chain-state feeds
Out of our scope
The answer lives in the real world. Still requires an oracle.
Who wins the election?
What is tomorrow's CPI print?
Did a specific event occur?
Sports outcomes
Arbitrary real-world resolution
Price of an off-chain asset

The chain is not trying to replace oracles for everything. It removes them for the class of markets where the settlement condition is a function of on-chain chain state. That class is larger than it sounds.

vs. Traditional Oracle

Why chain-native settlement is different

Property
Settlement source
Attack surface
Latency
Dispute mechanism
Verification
Methodology
Chainlink / UMA
External data feed
Feed manipulation, front-run
Minutes to hours
Dispute game / staking
Trust the reporter
Opaque or operator-set
Markovian
Block data
Chain attack only
Next block (~60s)
None needed
ZK proof, verify from block
On-chain, auditable

The ZK proof means you do not need to trust the miner, the data provider, or a dispute resolution committee. The computation is verified before the block is accepted. Settlement is as final as the block.

The methodology is not proprietary. The HMM transition and emission matrices are published in the whitepaper. The input signal construction is documented. Any node can reproduce the computation from the same inputs and verify the proof independently.

Get Involved

Open to developers, researchers, and builders

The protocol is live. The chain is running. The codebase is open. If you want to build on this primitive, a new chain, a new application, a fork with a different computation graph, start here.

Reach out directly: [email protected]
Read the spec: whitepaper  ·  ZK spec  ·  the standard
Follow the chain: explorer  ·  leaderboard

Read the Whitepaper SSRN Preprint 6920838 Start Mining API Reference