What is Myelin?¶

Myelin is an experimental CKB-isomorphic session runtime for finite Cell execution. It runs high-throughput Cell transitions off-chain, keeps them finite and typed, and emits evidence that can be projected toward CKB-style transaction contexts — with a future court path that lets a single disputed chunk be adjudicated by a CKB-VM-style verifier on the L1.

This page explains the positioning, what Myelin deliberately is not, and the claim ladder that any Myelin evidence has to climb.

The short version¶

Myelin is a CKB-style isomorphic session runtime: a finite off-chain
Cell ledger with typed conflict scheduling, deterministic VM
verification, and a future CKB-style court path for disputed chunks.

A more careful wording for public demos:

Myelin currently uses selectable closed-validator finality for session
benchmarking and pressure testing. The CKB-style projection and future
court path is what keeps it aligned with CKB semantics.

Both phrasings matter. The first is what Myelin aims to be; the second is what it is today.

What Myelin is¶

Property	What it means in Myelin
CKB-isomorphic	Same Cell mental model, same Molecule encoding, same syscall surface.
Finite Cell session	State is a finite set of Cells inside one session; no unbounded global state.
Typed conflict scheduling	CellDAG scheduler uses typed conflict hashes + read/write domains, not fee markets.
Deterministic CKB-VM verification	Scripts run in a RISC-V-based VM with the same determinism contract as CKB.
Selectable finality	Static closed committee or Tendermint-style weighted precommit, chosen per session.
CKB-style projection	Every chunk ships a `CkbProjectionReport` showing whether it's projectable.
Single-chunk court path	One disputed chunk is CKB-VM-verifiable on the L1; interactive bisection is a fallback.
Reference workload	The Teeworlds-on-CKB replayer is the canonical pressure test.

What Myelin is not¶

These are claims about the design, not missing benchmarks — the protocol surface makes them structurally out of scope:

Not a CKB full-node fork. Myelin does not import or sync the CKB client. It re-implements the parts of CKB it needs (Cell, CellTx, VM, syscalls) in its own workspace.
Not a new L1. Myelin does not run its own consensus on a independent network. Finality is a closed committee or Tendermint BFT — explicitly not Nakamoto PoW.
Not a permissionless L2 today. The static-committee and Tendermint engines both assume a known validator set. Until the L1 court path is implemented and exercised, Myelin should not be marketed as permissionless L2 security.
Not a general smart-contract platform. Myelin optimises for bounded sessions that produce challengeable settlement artefacts, not for arbitrary on-chain dApps.
Not a microsecond matching engine. Sub-millisecond matching, FPGA paths, and global public order books are outside the design.

The full positioning case (with use-case tiers) lives in the upstream MYELIN_USE_CASE_POSITIONING.md. The TL;DR:

%%{init: {
  "theme": "base",
  "themeVariables": {
    "primaryColor": "#A5B4FC",
    "primaryTextColor": "#1E293B",
    "primaryBorderColor": "#4F46E5",
    "lineColor": "#6366F1",
    "secondaryColor": "#C7D2FE",
    "tertiaryColor": "#C7D2FE"
  }
}}%%
flowchart LR
    subgraph IN["In scope (Tier 1)"]
        I1["Game sessions<br/>(Teeworlds)"]:::ok
        I2["Industrial IoT metering"]:::ok
        I3["RFQ / market-maker<br/>settlement sessions"]:::ok
        I4["Streaming payments"]:::ok
        I5["AI agent service receipts"]:::ok
    end
    subgraph VIA["Viable (Tier 2 — needs more)"]
        V1["Cross-org IoT"]:::warn
        V2["Batch auctions"]:::warn
        V3["Supply-chain receipts"]:::warn
        V4["Usage-based billing"]:::warn
        V5["Small MP tournament economy"]:::warn
    end
    subgraph OUT["Out of scope (Tier 3)"]
        O1["HFT matching engine"]:::no
        O2["Global public order book"]:::no
        O3["Unbounded MMO world state"]:::no
        O4["Raw sensor firehose as Cells"]:::no
        O5["Day-1 permissionless validator network"]:::no
    end
    classDef ok  fill:#C7D2FE,stroke:#7C3AED,color:#1E293B;
    classDef warn fill:#A5B4FC,stroke:#D97706,color:#1E293B;
    classDef no  fill:#C7D2FE,stroke:#DC2626,color:#1E293B;

The claim ladder¶

Every Myelin artefact has to climb a four-tier claim ladder. The ladder is deliberate: each tier requires a specific report to exist and to verify.

no projection report      -> designed to stay close to CKB semantics
successful projection     -> projectable into a CKB-style transaction/context
court bundle              -> executable disputed-chunk input shape
future exercised court    -> CKB-aligned adjudication path

Read Claim ladder for the exact artifacts that establish each tier and what would have to be true for Myelin to advertise a higher one.

How Myelin compares to other L2 patterns¶

Dimension	Optimistic rollup	ZK rollup	Sidechain	Myelin (today)	Myelin (target)
Validity model	Fraud proof	Validity proof	Independent consensus	Closed committee	Single-chunk CKB court
Dispute granularity	Instruction-level bisection	Per-tx validity	N/A	Whole chunk	Whole chunk
Asset custody	L1 bridge	L1 bridge	L1 bridge	L1 Cell lock + commit	Same
Finality	L1 finality + challenge period	L1 finality + proof time	Sidechain finality	Committee certificate	Court-verified
CKB-style projection	N/A (EVM-shaped)	N/A (EVM-shaped)	Optional	Yes, every chunk	Yes
Open permissionless	Yes	Yes	Yes	No	Planned

Myelin optimises for bounded, challengeable settlement artifacts, not for global public ordering. If your problem is "I need to settle 10k game sessions per second and challenge a disputed frame on CKB later," Myelin is structurally appropriate. If your problem is "I need a global order book with permissionless validators," it isn't.

A note on maturity¶

Myelin's current evidence boundary is honest about what is and isn't proven:

✅ A simple CellTx runs through Myelin's CKB-strict VM path and produces a CKB projection report with no deviation flags.
✅ The Teeworlds replayer binary runs end-to-end through the VM probe.
✅ Both finality engines (static committee, Tendermint) produce the same session ID, CellTx commitments, scheduler commitment, and state roots on a single-validator or quorum-validator fixture.
✅ The CKB devnet smoke test deploys DA-anchor and settlement CellScript carrier verifiers and proves live type-script execution on the parent CKB devnet.
⚠ The single-chunk CKB court verifier is not yet implemented on mainnet — the bundle is a self-contained, deterministic input ready for one.
⚠ Permissionless validator entry is not part of the current design.

That boundary is explicit on purpose. See Evidence paths for the full inventory.