Claim audit

P_M = min(|M|/500, 1) is an input to StationarySolver.

Prediction

Injecting transactions raises P_M and changes the residual of the next block.

Inside → outside

The node reads mempool.size, normalizes by 500, and passes it as a coupling.

Outside → inside

External transactions (faucet, wallet, simulated actors) increase |M| before the next solve.

Closure

Network → mempool → pressure → solve → block → (fees drain mempool) → next pressure.

If ablating λ_mempool leaves residual unchanged under load, the coupling is decorative.

Three independent mechanisms.

Prediction

A block can be stationary yet unfinalized; fork-choice uses cumulative residual; finality uses 2/3 stake.

Inside → outside

Solver admits a candidate; BFT votes separately; no code path equates the three.

Outside → inside

Explorer shows residual, cumulative residual, and finalized height as distinct fields.

Closure

Finality votes are recorded after commit and do not rewrite the residual.

If finality required stationarity or vice versa, the separation would be false.

External submission accepted claimed residual without rerunning the solver.

Prediction

Every accepted block is independently recomputed: continuity, merkle, residual, threshold, signatures.

Inside → outside

verifyStationaryEvidence() is called on every compose, not only on gossip.

Outside → inside

The Verify body re-runs the same function on the tip without access to miner search.

Closure

A failed check marks verified=false; it does not silently become canonical truth.

The original TS POST /blocks/submit path still exists in the source repo — territory, not this node.

Rust joint_residual uses raw u64 hash minus difficulty, almost never < 1e-7.

Prediction

Faithful territory residual remains huge; dimensionless canonical residual can fall below threshold.

Inside → outside

Both quantities are computed from the same header. Consensus uses canonical R.

Outside → inside

Every block header publishes both residuals. Experiments can plot them independently.

Closure

This is a rebuild decision, not a claim that the old solver already worked.

Calling the old residual 'solved' would falsify the audit.

SMT commits accounts, UTXOs, contract storage — not validators, DEX, mempool, peers.

Prediction

stateRoot changes when balances, pool reserves, the BTC tip, the ETH tip, or WASM storage change. Those inputs ride in the block. A peer disconnect does not.

Inside → outside

stateRoot merkle-izes account leaves, pool reserves, the BTC tip, the ETH tip, and WASM storage. The block carries the evidence a second body needs to reach that root. Peers stay outside it.

Outside → inside

Wholes panel lists committed vs operational vs observed vs whole.

Closure

A light node can prove an account or a pool reserve against stateRoot without the peer table.

If a peer disconnect changed the root, or a swap that was included did not, the commitment would be wrong.

Expensive solve, cheap re-verify.

Prediction

Recomputing residual from (header, txs, pressure) matches claimed R without nonce search.

Inside → outside

Discovery iterates nonce. Verification evaluates once.

Outside → inside

The Verify page, Mobile page, and Android client run the cheap path only.

Closure

Agreement is a measurement on the tested path, not a proof of every transition.

If verification required the same iteration budget, the split would be false.

Groth16/RISC Zero bind residual/threshold/block hash — not Ω_t → Ω_{t+1}.

Prediction

No block on this node carries a Groth16 proof of the full transition.

Inside → outside

successor is the transition. The stationarity relation on the header is not a Groth16 of that function, and it is not a stand-in for it.

Outside → inside

Research page states the limitation in the same words as the source LIMITATIONS.md.

Closure

We do not advertise a proof we cannot verify.

Shipping a fake 'valid: true' badge would falsify this claim.

Reward = base · min(1/(R+1e-6), 1), with coinomics quality.

Prediction

Lower residual yields higher coinbase, saturating at the base reward.

Inside → outside

composeBlock computes quality from canonical R.

Outside → inside

Block pages show coinbase next to residual. Library shows the coinomics curve.

Closure

Residual participates in acceptance, reward, and (via history) fork-choice.

A constant coinbase would falsify the coupling.

Bidirectional model is the claim to test, not a decoration.

Prediction

Each organ has at least one traced arrow in both directions, or is marked one-way.

Inside → outside

Solver emits blocks. Wallet emits signed txs.

Outside → inside

The browser recomputes R from the published header and compares it to the claim.

Closure

The loop page stores G(I) vs I(E) per height. Agreement there is not universal closure.

A marketing diagram with no event stream would fail this claim.

Disagreement is evidence.

Prediction

This node publishes where it matches the repo and where it rebuilds.

Inside → outside

Address derivation SHA-256(pubkey)[..20] matches both stacks.

Outside → inside

Header hash here commits merkle, state root, nonce, residual, and pressure.

Closure

The difference is listed in the audit, not papered over.

Pretending byte-identity with the Replit miner would be false.