Skip to content

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. Until a tier is reached, the corresponding claim is not made.

This page walks the ladder once, top to bottom.

The four tiers

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

%%{init: {
  "theme": "base",
  "themeVariables": {
    "primaryColor": "#A5B4FC",
    "primaryTextColor": "#1E293B",
    "primaryBorderColor": "#4F46E5",
    "lineColor": "#6366F1",
    "secondaryColor": "#C7D2FE",
    "tertiaryColor": "#C7D2FE"
  }
}}%%
flowchart TB
    T0["Tier 0<br/>Designed to stay close to CKB semantics"]:::t0
    T1["Tier 1<br/>Projectable into CKB-style transaction/context"]:::t1
    T2["Tier 2<br/>Executable disputed-chunk input shape"]:::t2
    T3["Tier 3<br/>CKB-aligned adjudication path"]:::t3

    T0 --> T1 --> T2 --> T3

    classDef t0 fill:#C7D2FE,stroke:#6366F1,color:#1E293B;
    classDef t1 fill:#A5B4FC,stroke:#4F46E5,color:#1E293B;
    classDef t2 fill:#C7D2FE,stroke:#7C3AED,color:#1E293B;
    classDef t3 fill:#C7D2FE,stroke:#7C3AED,color:#1E293B;

Tier 0 — Designed to stay close to CKB semantics

Claim. The runtime is designed to keep its transitions Cell-shaped and CKB-compatible. No measurement, no projection — just an architectural claim.

Evidence. This documentation. The architecture pages describe the runtime's structure; the concept pages describe the CKB mental model Myelin borrows from.

Boundary. Tier 0 is the floor. It is not sufficient to claim CKB alignment; it's sufficient to claim that the design intends to be CKB-aligned.

Tier 1 — Projectable into a CKB-style transaction/context

Claim. A specific transition has been measured and projected; the projection report either confirms ckb_projection_possible: true or lists explicit deviation flags.

Evidence. A MyelinExecutionReport paired with a CkbProjectionReport. The two together prove:

  • The transition ran in Myelin's CKB-VM-style verifier with a measured cycle count and exit code.
  • The transition is projectable into a CKB-style transaction with a deterministic CKB transaction hash.

What it doesn't prove. The transition is projectable, not projected. The bytes haven't been submitted to CKB. The court verifier hasn't replayed them. The projection is a proof-of-shape, not a proof-of-validity on a live chain.

How to reach it. Run cargo run -p myelin-cli -- celltx simple-report and check the output JSON has semantic_profile: "ckb-compatible" and ckb_projection_possible: true.

Tier 2 — Executable disputed-chunk input shape

Claim. A specific disputed chunk has been packaged as a self-contained input that a CKB-VM-style court verifier could consume.

Evidence. A court-bundle JSON that re-verifies successfully:

vm_profile            -> matches declared profile
spawn/ipc requirement -> consistent
payload_hash          -> matches chunk_payload
molecule_tx_hash      -> matches ckb_molecule_tx_bytes
projection_hashes     -> matches projection_report
challenge_payload_hash-> matches challenge_payload
signature_hashes      -> matches committee evidence
signer_ids            -> valid committee certificates
committee certificate -> quorum weight present
court_verifiable      -> single-chunk verification possible
semantic_profile      -> matches payload
ckb_projection_possible-> matches projection report
unsupported_features  -> listed correctly
semantic_deviation_flags -> listed correctly
l1_court_implemented  -> false (explicit)
challenge_window_consistent -> with settlement intent if present

(That's 16 distinct assertions; the verify command runs them all.)

What it doesn't prove. The CKB court verifier exists yet. Tier 2 is the input shape to that verifier — it's the artifact that would be adjudicated, ready and waiting.

How to reach it. Run the session court-bundle commands:

myelin-cli session court-bundle  --session <commit>  --chunk-index 0  --out reports/bundle.json
myelin-cli session verify-court-bundle  --bundle reports/bundle.json  --out reports/verify.json

The verify report must report valid: true with all 16 assertions passing.

Tier 3 — CKB-aligned adjudication path

Claim. A CKB court verifier type script exists, is deployed on a live CKB chain, and has actually replayed a bundle to a verdict.

Evidence. A CKB transaction hash on mainnet or testnet, with:

court verifier type script deployed on the target chain
disputed chunk bundle submitted as input
on-chain verdict emitted (accept or slash)
slashed bond visible on chain

What it doesn't prove (yet). That Myelin is a permissionless L2 in the public-validator sense. Tier 3 is the adjudication tier — it's about whether disputes can be resolved on L1, not about who can be a validator.

How to reach it. This requires:

  1. A court verifier type script written in CellScript (or hand-rolled C/Rust).
  2. The verifier deployed on CKB mainnet (or a testnet that counts for the claim).
  3. A bundle submitted to the verifier.
  4. The verifier emitting a verdict.

This is future work. The Myelin kernel can produce all the inputs today; the L1 court path is what closes the loop. Until then, every Myelin claim sits at Tier 2 or below.

Where each Myelin artefact sits today

Artefact Tier Why
simple-report.json (trivial CellTx) 1 Has an execution report + projection report.
session-court-bundle.json + verify-court-bundle.json 2 All 16 assertions pass; bundle is self-contained.
session-da-anchor-package.json + verify-da-anchor-package.json 2 (anchor) Package is verified; submission is dry-run only.
session-settlement-package.json + verify-settlement-package.json 2 (settlement) Same as above for settlement.
Devnet smoke myelin-ckb-devnet-smoke-v1 2 (with live CellTx execution) Live carrier submissions on a parent CKB devnet, including tampered-carrier rejection.
Real CKB court verdict on mainnet 3 Not yet. Requires court verifier deployment.

The discipline

The ladder's discipline is simple:

  • Don't claim a tier you haven't reached.
  • Don't skip tiers; each one is built on the previous.
  • Don't hide the boundary — every report carries explicit flags that say which tier it sits at.

The Myelin CLI's readiness reports are designed to make this hard to fudge:

readiness {
  semantic_profile:              "ckb-compatible"   // tier >= 1 if ckb-compatible
  ckb_projection_possible:       true               // tier >= 1
  l1_da_published:               false              // tier < 3 for DA claim
  l1_court_implemented:          false              // tier < 3 for court claim
  production_submission_ready:   false              // tier < 3 for end-to-end
  end_to_end_production_ready:   false
  production_blockers:           [...]              // explicit
}

Anyone who reads the report sees exactly which claims are established and which aren't.

Where to go next