Amadeus Protocol

Audit date · 2026-11-01

Audits are prepared by the site owner with a self-hosted (local) AI, checked against the exact commit above, and hand-edited before publishing. This is not financial, legal, or investment advice. Read it as one person's notes on one commit at one date. Do your own research.

This is a point-in-time review of code quality and tokenomics at the commit above, not a safety certification. Projects can change after this date. Check for newer releases or a newer audit before relying on these scores.

Useful-PoW L1 with Freivalds-sampled work verification and an epoch validator set rebuilt from vault stakes. The money rules are pure integer arithmetic, but 25% of paid emission goes to a single address and the live supply cannot be produced by the audited emission formula.

🪙 AMADEUS PROTOCOL (AMA) — Code & Tokenomics Audit

Repo: amadeusprotocol/node at commit ab67f03a6ffeb96acd4bde4d962bcf19e7b58164 (Elixir node, Rust consensus NIF)
Date: 2026-11-01 Scope: code review of the money path (emission curve, coin create/mint, treasury tax, vaults, validator-set rules, key storage), tokenomics, live network numbers from nodes.amadeus.bot/api/chain/stats fetched 2026-11-01. The Freivalds sampler was read as a statement, not cryptanalytically re-proved. Tests read, not executed.

🎯 Verdict: MIXED (code 7.0 / tokenomics 4.0)

Amadeus is an L1 whose "mining" is sampled useful computation: solvers submit a matrix-multiplication solution, verified by three-vector Freivalds against a committed verifier random, with BLS aggregate signatures and a validator set rebuilt each epoch from vault stakes. The rules in the audited code are clean arithmetic: the emission curve is a pure integer function, the 25% network tax is funded from the leftover pool rather than minted separately, the bridge is proof-gated against stored remote roots, and the validator roster is rule-based (every vault validator with ≥1M AMA stake plus the top 33 solvers), not a curated list.

It lands in MIXED on three tokenomics facts. The network tax is 25% of paid solver and vault emission — above the 10% band in this corpus. The tax lands in the balance of a single BLS address, TREASURY_DONATION_ADDRESS, which is also the hardcoded cold admin of PRIME, an admin-mintable symbol; no timelock or multisig exists on-chain. And the observed supply cannot be produced by the audited code: the curve this repo contains sums to 30,162,597 AMA through epoch 892, while the live chain reports 709,483,635.15 AMA total supply — 95.7% of the money on the chain did not come from this repo's emission formula. The published claim that the 1B cap is "emitted entirely through useful proof of work with no premine" is not reproducible from the audited code at this commit; the only code path that could produce the observed state is the coin create_and_mint admin path or a supply rule living outside this repo. On the code side: unsafe lifetime transmutes through the consensus NIF, the work acceptance is probabilistic (error ≤ 1/8 per accepted solution), the node key is stored plaintext, and the repo has no CI.

🔑 At a glance

Live numbers (fetched 2026-11-01, height 89,275,658): circulating 597,280,526.15, total supply 709,483,635.15, vault-locked 112,203,108.99, burned (fees) 2,274,223.80, 121 entries in the validator set, emission for the current epoch 188,952.39.

🔗 Links & sources

🧮 Risk score inputs

code deductions, one at a time:
  1.0 -> 2.0 -> 3.0
tokenomics deductions, one at a time:
  3.0 -> 5.0 -> 6.0

code_score       = 10 − (1.0 + 0.5 + 1.0 + 0.5)  = 7.0
tokenomics_score = 10 − (3.0 + 2.0 + 1.0)         = 4.0
risk_score       = min(7.0, 4.0)                  = 4.0

Band check: 4.0 sits in 3.0-7.9 = MIXED (vocabulary from AGENTS.md). No instant-HIGH-RISK override fired: no claim the code contradicts outright was proven — the supply-origin item is an arithmetic unreachability, stated as such.

✅ Verified good (code)

• Emission is a pure integer function per epoch, no floating point, checked_mul / saturating_sub throughout the money path (25 guarded arithmetic sites in the consensus crate).
• The tax is funded from the undistributed leftover, not minted on top (epoch.rs:245 comment and code agree).
• Coin create and mint are capped (mint_amount_exceeds_cap at coin.rs:180 and :234).
• Bridge: proof-gated against stored remote state roots, once per nonce, and the escrow address is not a valid BLS key so nobody can sign for it (bridge.rs:1-25).
• Validator set is rule-based each epoch: all vault validators with stake ≥ VALIDATOR_MIN_STAKE (1M AMA) plus top 33 solvers, reshuffled; vault-to-validator changes pass through a pending-epoch queue (lockup_vault.rs:575, 643).
• Tests exist for both curves (golden vectors at epoch.rs test asserts), mix.lock present, release signing script present.

🔍 Open questions (not scored)

• Where the 679.3M AMA above this repo's curve came from: the BSV-side hub contract, a supply rule outside node, or the bootstrap vault path. Needs the launch record to answer.
• Live emission_for_epoch (188,952.39) is 249.2 AMA above the formula value for epoch 892 computed exactly in integers (188,703.16); the difference is unexplained in the audited code — possibly carried pool amounts.
• Tests were read, not executed; the Freivalds sampler not cryptanalytically re-proved; PRIME-to-$AMA conversion (TGE) unpublished as of 2026-10-16 per third-party reports.
• 25% tax band value is inference from published pairs; the rubric skill was not available in this pass.

Files we opened

ex/native/rdb/src/consensus/bic/epoch.rs:13-15

ex/native/rdb/src/consensus/bic/epoch.rs:32-36

ex/native/rdb/src/consensus/bic/epoch.rs:41-47

ex/lib/api/api_chain.ex:141-161

ex/lib/consensus/models/entry_genesis.ex:157-171

ex/native/rdb/src/consensus/bic/prime.rs:25-45

ex/native/rdb/src/lib.rs:35-36

ex/native/rdb/src/consensus/bic/sol_freivalds.rs:68-110

ex/config/runtime.exs:61-79

ex/native/rdb/src/consensus/bic/lockup_vault.rs:575

ex/lib/api/db_entry.ex:174

Every file listed here was opened by us during the audit. Nothing here is a guess.

Back to Projects