Integrity board

The ten largest subnets take roughly two-fifths of network emissions. Top-10 subnets take 40.05% of emissions on the chain read, up from 33.2% at an earlier read. uniform alpha emission x alpha price / total = standard dTAO emission-value share

◆Established ● on-chain Price-weighted emission share across all subnets. The share is wired, so a material move routes to review.

Stake on most covered subnets is concentrated: on the median subnet a single coldkey holds a majority of the stake. Across the covered subnet set on 2026-09-18 (block 9,088,526) the median subnet has a single coldkey holding 57.26% of its stake, the worst holds 97.84%, and 10 of 15 have one coldkey with an outright majority. Per-subnet capture moves in both directions: Templar (SN3) has gone from 95.6% to 98.32% while Targon (SN4) has fallen from 88.3% to 52.35%. Per-subnet metagraph stake grouped by coldkey, then aggregated across the covered set. Grouping by coldkey rather than hotkey is the step that matters: one operator running many hotkeys reads as many holders otherwise.

◆Established ● on-chain Distribution rubric D0 on ongoing concentration despite a mining-only launch. Positive controls exist (Ridges, Score/SIRE around 29% top-1) and are recorded in the flags.

Bittensor's core code is public on GitHub, but a third of the org's own repositories carry no licence file and most have not been pushed in six months. Re-read 2026-09-18: 35 non-fork public repositories in github.com/opentensor (50 public, 15 forks), 11 with no licence file and 33 not pushed in 180 days. Core code is open (MIT); the Subtensor runtime is public - we read the PR #2781 Rust diff directly this cycle. Caveat: no formal security audit of the core chain. Confirmed the public MIT repos and read core Subtensor source directly.

◆Established ● on-chain We read the Subtensor source directly for the emission-mechanism check. The assurance gap is the absence of a formal security audit of the core chain.

TAO is required on-chain to stake, to register miners and subnets, and to enter each subnet's alpha market. Staking, registration-burn, and the dTAO subnet-token mechanism are all on-chain and load-bearing for participation. Confirmed the staking/burn/dTAO mechanisms on-chain.

◆Established ● on-chain Re-checked 2026-09-18: staking and dTAO are live and measured in our own index, with 129 subnets and per-subnet coldkey stake enumerated. The burn leg is the registration cost, a runtime parameter, and the runtime is watched at spec 466.

Lium's daily provider payouts on chain match the gross billing it publishes, so its self-reported revenue is backed by money that actually left the subnet. Lium reports $1,295,273 of billing for September 2026 ($15.5M/yr). Its payer coldkey's SN51 stake transfers at the documented 17:00 UTC payout averaged 136 TAO/day across 7/7 days to ~70 recipients, which at the run's TAO price and the documented 95% provider share implies $15.9M/yr of billing, within 3% of the reported figure. Lium's monthly billing was $437,737 in September 2025, so 3.0x year on year. Summed the payer coldkey's SN51 StakeTransferred amounts (TAO-equivalent) in the 25 blocks after 17:00 UTC for each of the last 7 days, valued at one TAO price, divided by 0.95, annualised, and compared with Lium's last full month of billing x12. The payer is identified by behaviour: on every day sampled (10 and 20 September, 1, 4 and 5 October 2026) it was the only coldkey making SN51 stake transfers at that time.

◆Established ● on-chain Establishes that the billing is real money paid out, not that the renters are outside customers. Renters pay off-chain, so whether they are external businesses or other subnets' miners spending emission income cannot be read from chain. The alpha paid out comes from SN51 emission routed to Lium's payout key (recorded in the next finding), not from purchases with renter revenue.

Lium pays its providers' rental share from SN51 emission routed to its own payout key, and spends renter TAO buying and burning SN51 alpha. Lium's validator config routes every burn slot to UID 47 ('All burn emission slots route to UID 47 (Datura payout key)', expected coldkey 5G694c15…), and on chain UID 47 is owned by the payout coldkey and takes most of SN51's miner incentive (77.5% on 6 October, 84.4% on 7 October 2026, moving with the unrented share). That coldkey's SN51 stake rises each tempo and falls at the 17:00 UTC payout, so the provider share is paid from emission. Small TAO transfers from many wallets reach the owner coldkey 5FqACM…, which bought and burned SN51 alpha with 3,737.7 TAO (9 Sep 2026), 701.0 TAO (18 Sep) and 1,230.5 TAO (25 Sep). Read UID 47's coldkey and incentive from the metagraph; sampled the payout coldkey's SN51 stake hourly across a payout to separate per-tempo emission from the daily transfer; bisected the owner coldkey's nonce over 36 days and decoded each block where it moved.

◆Established ● on-chain Lium's provider docs call the pool 'burn' ('Distributed equally across designated burner UIDs'); the deployed config sends it to Lium's payout key, so nothing in that pool is burned. That the TAO arriving at the owner coldkey is renter top-ups is consistent with Lium accepting TAO (docs.lium.io, lium fund) but the senders are not individually attributed. Card payments are not visible on chain. Burns are lumpy: the 9 September burn spent a balance accumulated through August, so a single 30-day burn total overstates the steady rate; the refresher records the owner's actual TAO inflow over the window and the share of payout-key alpha it would buy back.

Subnet participant rewards are overwhelmingly funded by TAO emissions; measured demand revenue covers a small fraction of the minted budget. Emission budget 3,573 TAO/day (1.30M TAO/yr), derived from the on-chain TotalIssuance delta. Revenue in tiers, read 2026-10-06: settled by a third party, $0.85-5.19M/yr (OpenRouter's token series for Chutes, SN64), 0.2-1.3% of the budget; the subnets' own current figures, $18.8M/yr (Lium's billing, corroborated by its on-chain provider payouts, plus Chutes' list-price usage and Bitcast's dashboard), 4.7%; adding Targon's year-old $10.4M claim, a generous ceiling of $29.2M/yr, 7.4%. Participant rewards are overwhelmingly inflation-funded on every tier. Emission from the TotalIssuance delta, not BlockEmission storage (reads 1 TAO/block against ~0.496 actual). Revenue kept in tiers that are never blended: measured (a third party settles it), self-published (the subnet's own current figure, flagged where our chain read corroborates it), and a generous ceiling that adds stale claims. Ratios at one TAO price captured on the run.

◆Established ◑ Cross-checked Graded api because the binding measured half is OpenRouter's series. The self-published tier is the subnets' own figures; Lium's is tested against its on-chain payouts and matches, but gross billing does not show whether renters are outside customers or other subnets' participants spending emission income.

Capital in DeFi on the Subtensor chain is largely subnet-token liquidity. $618,984,441 DeFi TVL on the Bittensor chain (DeFiLlama), re-read 2026-09-29 ($531,623,010 on 2026-09-24). The step from about $47M to $514M on 19 September is DeFiLlama adding a 'Bittensor dTAO' protocol (subnet-pool liquidity) to the chain total, not new capital. On the previous basis the figure was $42,824,796 on 2026-09-24. DeFiLlama chain-TVL sum, refreshed by scripts/refresh-primary.ts into meta.primary_data.

◆Established ◑ Cross-checked Recorded with its limit stated: chain TVL is capital in DeFi on Subtensor, not the emission economy the review rates. Kept because subnet-token liquidity is the surface a holder transacts against.

TAO trades across a broad set of exchanges, led by KuCoin, Binance and WhiteBIT. 72 tickers on CoinGecko, re-read 2026-09-24 (82 on 2026-09-18, when KuCoin, Binance and WhiteBIT led 24-hour volume). Read the ticker set and 24h volume live.

◆Established ◑ Cross-checked Structural framing only. The ticker count is wired; turnover is measured alongside it.

Bittensor is a live network with 129 subnets registered on-chain. 129 subnets registered on-chain, read from the Subtensor finney subnet set. Subnet count on-chain; market scale from CoinGecko.

◆Established ● on-chain Whether the subnets do useful work is not measurable on-chain (subnet_output flag).

Control over Bittensor's chain is centralised: one sudo key dispatches privileged operations and blocks come from authority nodes. Read off the runtime on 2026-09-18 (block 9,088,526): finney carries 28 pallets and none of them is a Triumvirate, Senate, Collective, Council, Democracy or membership pallet, while Sudo.Key holds 5DcSqBNqCmfdJZRGFSwwcRb2dZdJHZuKK8Tb1Gx8gbmF5E8s. Privileged operations, runtime upgrades included, dispatch through that key. On 2026-09-25 the chain showed that key is a 2-of-3 multisig: the address derives from signers 5GRCukV2rZmSVfJhAXoLjcrU1pMVCf2Ra1ydbiCFZdaQXDXo, 5E7RCRrPVS8TckCDjr92B5ciGziwz2kfvxe4URy3L7AgirGJ and 5FevFjov8435t5XC2MUSRpFYxtthE8pZy1toHpgAAia3ZphG at threshold 2, and the spec 468 setCode at block 9,117,748 executed through it with two approvals (via a Proxy.proxy from a 2-of-2 multisig that includes the sudo key itself). Bittensor's own validator documentation names the signers as Rao Foundation, which we record as their statement, since the chain shows accounts, not who holds them. Block production remains authority-based, and OTF has used a root override before (SN28). Stake is heavily captured per subnet: the median covered subnet has one coldkey holding 57.26% of its stake and 10 of 15 have an outright majority holder. dTAO added market allocation of emissions, and root control did not move with it. Runtime metadata plus Sudo.Key for the governance surface, per-subnet metagraph stake grouped by coldkey for the concentration leg. Both are chain reads; neither is taken from documentation.

◆Established ● on-chain dTAO moved emission allocation to a market mechanism, and root control did not move with it. The sudo key is wired, so a change in who holds it routes to review.

Emissions fund the overwhelming majority of subnet participant rewards, even with every self-reported revenue figure added. The network mints 3,573 TAO/day (1.30M TAO/yr) to pay participants, derived on-chain, about $397M/yr at the run's TAO price. Revenue a third party settles is $0.85-5.19M/yr. The subnets' own current figures sum to $18.8M/yr (4.7%), most of it Lium, whose billing its on-chain payouts corroborate. Even with Targon's stale claim added, $29.2M/yr (7.4%): emissions fund the overwhelming majority of participant rewards. Compared the on-chain minted budget to each revenue tier: measured floor and ceiling, self-published total, and the generous ceiling.

◆Established ◑ Cross-checked Some subnets have paying products, and Chutes' OpenRouter traffic is paid demand; the network-level economics are still emission-subsidised. The ratio is not the banded figure, because revenue is billed in dollars and the budget minted in TAO; the band sits on the minted budget, which moves only if the protocol does.

The canonical VIRTUAL token on Ethereum has renounced ownership, no upgrade path and a fixed 1,000,000,000 supply. Canonical L1 VIRTUAL (0x44ff86, Ethereum): ownership renounced (owner = 0x0), not upgradeable (no EIP-1967 impl slot), totalSupply fixed at exactly 1,000,000,000. The Base token is bridge-minted via the canonical Base Standard Bridge (0x4200...0010), not a discretionary team mint. Read L1 owner() (0x0), totalSupply (1B), EIP-1967 slot (zero); confirmed the Base contract is an OptimismMintableERC20 whose minter is the Standard Bridge.

◆Established ● on-chain The Base token is minted only by the canonical Base Standard Bridge. The control concern sits on the protocol contracts (core_contract_control), not the token.

On the Base leg, VIRTUAL holdings outside infrastructure addresses are moderately concentrated, and none of the raw top ten is identifiable as an insider. Measured on the BASE leg on 2026-09-18 across 1,201,365 holders: the top ten hold 46.73% of supply raw and 21.05% once infrastructure addresses are excluded, with 26.63% of supply in infrastructure and 0% of the raw top ten identifiable as insider. 5 of the top ten are unlabelled. SCOPE: Base only. VIRTUAL also has live balances on Ethereum and Solana, and summing legs would double-count any holder on two of them, so this is one leg named as such rather than a network total. Top-10 holders with infrastructure addresses classified and excluded from the numerator while their balances stay in the denominator. One chain leg, scope stated: the cross-chain merge that would rate all three is a build, and a partial sum would be worse than a named partial view.

◆Established ● on-chain Measured 2026-09-18 on one chain leg, with the scope named. A three-leg merge remains a build, because a naive sum double-counts any holder present on two legs.

VIRTUAL holders cannot be frozen or blacklisted at the token layer: the L1 token is renounced, non-upgradeable and has no blacklist or pause path. L1 token ownership renounced and non-upgradeable; standard ERC-20 with no blacklist/pause path. Confirmed renounced ownership and no upgrade path; no blacklist/freeze mechanism.

◆Established ● on-chain Platform-layer censorship is a separate question (platform_censorship).

The VIRTUAL token and the AgentNftV2 registry carry verified source on the block explorers, and the protocol-contracts repository is public. AgentNftV2 and the VIRTUAL token show verified source on the block explorers; the protocol-contracts repo is public. Confirmed contract verification on-chain and the public repo set.

◆Established ● on-chain Open-source rubric about O2 on contracts (verified, public). The protocol-contracts repository carries no licence file (GitHub API, 2026-09-23), and the GAME runtime is not fully open, so platform transparency is limited. Source verification is re-read daily by the control-surface watch, which also reads back the contract name as a control.

Protocol fees are measurable on-chain and falling: the 30-day run-rate sits well below the trailing year, and the protocol keeps only part of them. $15.71M trailing-year and $74.6M all-time on-chain fees (DeFiLlama virtual-protocol adapter), re-read 2026-10-07. Since 5 October DeFiLlama books only the treasury's share as revenue: $4.85M trailing-year and $41.7M all-time, with the rest of the agent-token trading tax paid to agent creators (the adapter applies the whitepaper's 30/70 split to the measured tax). The 30-day fee window is $519,816, annualising to about $6.3M, so the trailing year is still carrying fees the current run-rate is nowhere near reproducing. DeFiLlama indexes the multi-stream protocol take across chains - Base legacy streams + cbBTC prototype + tax-manager sentient outflows + Ethereum + Solana + the 1% agent-trading tax - replacing ecosystem/treasury transfers with the tax-manager outflow to avoid double-counting. Our own naive getLogs self-index was not tractable, so we cite DeFiLlama's open adapter. Since PR #9918 (5 Oct 2026) the adapter separates the creator share of the agent tax as supply-side revenue; that share is computed from the whitepaper ratio, not traced to creator wallets.

◆Established ◑ Cross-checked Fees are about 97% below the January 2025 peak. Agentic GDP is a different, off-chain aggregate and is not protocol revenue (agentic_gdp). Banded on trailing-year FEES, which can fall, rather than the all-time total; moved from the revenue series on 2026-10-07 when DeFiLlama's PR #9918 split creator payouts out of revenue.

VIRTUAL trades across hundreds of venues, both centralised exchanges and on-chain pools. 457 tickers; real CEX plus Aerodrome DEX depth; turnover runs at a 7-day median of ~8.2% of market cap; ~89% below ATH. Volume skews to mid-tier venues (HTX, Biconomy, Toobit) alongside the majors. CoinGecko ticker list paginated to a real total, and turnover as the median of the trailing 7 daily volume/market-cap points. Refreshed by scripts/refresh-primary.ts.

◆Established ◑ Cross-checked Structural language only. The deep drawdown from ATH tracks the revenue decline. The ticker count is read paginated to a full total.

VIRTUAL supply is fixed and fee capture is live on-chain, but trailing fees are falling and the treasury keeps only part of them. $15.71M trailing-year and $74.6M all-time on-chain fees (DeFiLlama virtual-protocol adapter), re-read 2026-10-07. Since 5 October DeFiLlama books only the treasury's share as revenue: $4.85M trailing-year and $41.7M all-time, with the rest of the agent-token trading tax paid to agent creators (the adapter applies the whitepaper's 30/70 split to the measured tax). The 30-day fee window is $519,816, annualising to about $6.3M, so the trailing year is still carrying fees the current run-rate is nowhere near reproducing. Supply/renouncement read on-chain; revenue via DeFiLlama; allocation documented.

◆Established ● on-chain The token-economics half holds. The value question is whether agent demand recovers, with revenue about 97% below its January 2025 peak.

The launchpad operates at scale on-chain, measured by the graduated agents in its canonical Base registry. 1,233 graduated/Sentient agents in the canonical Base registry (live), a large multi-chain launchpad footprint. Registry totalSupply read live.

◆Established ● on-chain Scale is measured; dominance against rival launchpads is not. Whether the scale is productive is the thesis_productive_agent_economy row.

Most graduated Virtuals agents have gone quiet: the large majority trade under $1k of DEX volume in 30 days, and only a handful trade over $1M. Of 1,232 graduated agents (2026-08-10 index), 1,036 (84%) traded under $1k in 30 days (dead or near-dead), 514 (42%) had zero DEX trades, and just 3 (0.2%) traded over $1M/30d. The whole graduated set turned over $17.6M in 30 days. The live set has thinned since the June read: dead/near-dead 71% to 84%, and agents above $1M/30d 25 to 3. Per-token 30d DEX volume across the graduated registry; dead/near-dead = under $1k/30d, which includes agents with no DEX trades at all. See the survivorship metric_flag for the full breakdown.

◆Established ● on-chain The dead or near-dead share rose from 71% at the 2026-06-09 read to 84% while the graduated count barely moved (1,221 to 1,232), so the tail grew by agents going quiet rather than by new launches. Agents turning over $1M or more in 30 days fell from 25 to 3. The index measures trading liveness, not the work agents do.

Half of ATH supply is allocated to checker nodes and compute providers, vesting over time. Allocation as documented: 50% to compute providers (linear vesting). Read the allocation table; confirmed total supply on-chain (42B).

◆Established ◑ Cross-checked Aethir publishes the split only as an image (Tokenomics3D.png on its Token Distribution page, re-read 2026-09-23), so the allocation carries no quotable text. Total supply reads 42B on chain, which is the half of this row that can move. Insider concentration is recorded on insider_allocation.

Insiders hold a material, vested share of ATH: team, investors and advisors together carry close to a third of supply. Insider allocation is material: 12.5% team, 11.5% investors and 5% advisors, 29% between them, against 50% to checker nodes and compute providers and 6% to airdrop. Aethir's Token Distribution page publishes the split only as an image (Tokenomics3D.png), re-read 2026-09-23; the image carries the 5% advisors slice, which the Token Vesting table does not list as a separate row. Summed the team, investor and advisor allocations from the stored allocation table, then re-read Aethir's docs to check the table still stands. The docs render in JavaScript, so this was read in a browser and cross-checked against the .md rendition GitBook serves at the same paths.

◆Established ◑ Cross-checked Distribution rubric about D1: a notable insider allocation, vested, on a VC-backed project. Left manual on a 365-day cadence because a launch allocation does not drift. The stored allocation table sums to 100.17% because it carries 0.15% public sale and 0.02% pre-sale lines that the image does not show.

The ATH token has no pause, freeze or blacklist function and cannot be upgraded, so holders cannot be frozen at the token layer. Verified AethirToken source exposes an owner-only mint (exhausted at the MAX_SUPPLY cap) and an owner-only whitelist that pays out the contract's own balance; no pause, freeze, or blacklist function, and the contract is not upgradeable (no EIP-1967 slot). Enumerated the verified ABI (no pause/blacklist); confirmed non-upgradeable.

◆Established ● on-chain The owner's whitelist functions move only tokens the contract itself holds, so they cannot block a holder's transfer. Supply is recorded on hard_cap_enforcement. Network-layer censorship is outside this row: the runtime is closed-source.

Aethir publishes a handful of client-side repositories, none of them licensed, and the orchestration and indexer stack is not public. Seven own public repositories across two orgs, none of them licensed. AethirApp, the org Aethir's own docs send node operators to, holds two (checker-client, libyxrtc); AethirCloud holds five more (client-sdk-js, checker-client, HostAgent, metamask_demo, a profile repo) plus one fork. Not one carries an SPDX licence, so a reader has no right to use, modify or redistribute any of it. The orchestration and indexer stack is absent from both orgs, and only the ATH token contract is source-verified on Etherscan. Enumerated both orgs through the REST API for licence, fork status, stars and last push, after resolving which org Aethir itself points at from its own checker-node documentation.

◆Established ◑ Cross-checked Open-source rubric stage O0: the network core is not public. A verified token contract is not network transparency. Code published without a licence grants a reader no rights to use, modify or redistribute it. Open Source Transparency 2/15 is flagged for the monthly review, since published-but-unlicensed sits a notch above nothing published.

Holder value accrual is indirect: GPU operators receive about 80% of service fees, the protocol keeps about 20%, and nothing is shared with or burned for holders. The 80/20 split is measurable from the two DeFiLlama streams we already hold and it computes exactly: $36,058,888 protocol revenue against $180,294,425 gross fees, all-time, re-read 2026-09-18, which is 20.0% to the protocol and the rest to GPU operators. On-chain: 80% of service fees flow to GPU operators (supply-side), the protocol retains ~20%; there is no fee distribution to token holders and no buy-and-burn. Holder value accrual is indirect and thin. Measured the fee split (80% operators / 20% protocol) on-chain; confirmed no burn/fee-share mechanism.

◆Established ● on-chain Aethir's tokenomics page says value accrues to ATH and its holders as the ecosystem grows, without naming a mechanism. The split is measured from the fee and revenue streams.

The 42B ATH supply is enforced in code and fully minted: mint() cannot exceed the MAX_SUPPLY constant, totalSupply() already equals it, and the contract has no burn, so no more ATH can be minted. AethirToken declares MAX_SUPPLY = 42,000,000,000 ATH as a constant, and the owner-only mint() requires totalSupply() + amount <= MAX_SUPPLY. totalSupply() reads exactly 42B and there is no burn function, so the owner (a 2-of-3 Gnosis Safe, no timelock) cannot add supply. Its remaining power is an owner-only whitelist that pays out the small balance held by the contract itself. Read the MAX_SUPPLY constant and the mint guard in the verified source, confirmed the contract is not a proxy, and read totalSupply() and MAX_SUPPLY() by RPC (equal).

◆Established ● on-chain The contract exposes no cap() getter; the cap lives in the MAX_SUPPLY constant that the mint guard checks. Dilution comes from distribution of already-minted tokens (vesting and provider rewards) rather than new issuance. Aethir's MiCA whitepaper describes the supply as capped, but its PDF carries no readable text layer, so the statement cannot be quoted.

Aethir's headline revenue figure is gross developer spend; the protocol keeps about a fifth, and GPU operators receive the rest. The protocol's own take, the 20% fee cut, is $36,058,888 all-time on DeFiLlama's Revenue series against $180,294,425 gross fees, re-read 2026-09-18; the other 80% flows to GPU operators. Separated gross developer service-fee throughput from the protocol's 20% take; DeFiLlama Revenue confirms the protocol line.

◆Established ● on-chain Aethir's dashboard labels its headline Total Network Revenue, which is gross spend. Protocol revenue is the 20% cut, about $17.4M a year on DeFiLlama.

ATH trades on many venues, including large centralised exchanges, with steady daily turnover. 58 tickers, including centralised exchanges (Upbit, Bybit, LBank, CoinW, HTX); about 5.5% daily turnover of market cap; about 97% below the all-time high. Read the ticker set and 24h volume live.

◆Established ◑ Cross-checked Multi-venue access, stated in structural terms. The drawdown from the all-time high runs alongside the fall in service-fee revenue.

Aethir's gross service-fee revenue is on-chain and reproducible: our own index of the fee contract tracks the independent aggregators closely. Our own on-chain index of the AETHIR_CORE fee contract reproduces the trailing year within 4.1% of DeFiLlama ($78.7M against $82.1M), and matched DeFiLlama and Token Terminal to about 0.2% on calendar 2025. DeFiLlama's all-time gross fees read $180.3M (2026-09-18). Independent on-chain reconstruction of the service-fee stream.

◆Established ● on-chain The protocol's take is about 20% of this; see revenue_headline_basis. Our index can price only the window the stored price history covers, so the all-time dollar leg is DeFiLlama's to state.

About 69% of the 426.5M OLAS insider allocation was contract-locked at genesis, and every one of those locks had ended by July 2026. The 426.5M insider allocation was minted on 30 June 2022 to two addresses: 292,874,580 OLAS to the Sale contract 0xd1155408d58293be0743225bcde28b9fd0c12378 (block 15050749) and 133,625,420 OLAS to a Safe v1.3.0, 2-of-3, at 0x87cc0d34f6111c8a7a4bdf758a9a715a3675f941 (block 15050757). Sale can only pay out through claim() into veOLAS or buOLAS: 75 accounts claimed 181,002,829 OLAS into veOLAS locks of one to four years and 20 accounts 111,871,751 OLAS into buOLAS, all by block 15773895 (October 2022), so 68.7% of the insider allocation was locked from the start. The Safe later sent 42.8M into veOLAS locks for nine accounts, 14,812,100 into a TokenVestingManager (0xee99de3b4ab71f8f6bcdbabb758c066d3359b237, 22 schedules, all ending 7 June 2026, the Safe its sole admin with revoke rights), 1.30M into buOLAS and 1.27M into a Uniswap V2 pair. Every original lock had expired by 18 July 2026. Re-read 2026-10-07 (block 26138904): three of the 75 Sale claimants hold new veOLAS locks of 20.9M to September 2027 and May 2028, one Safe-funded account 0.25M to November 2026, buOLAS holds 55.1M of which 52.7M is revoked and permanently locked, the vesting manager holds 3.04M vested but unclaimed, and the Safe holds 52,736,093 OLAS with no lock. Summed the genesis mints and matched them to the insider allocation, read the Sale contract's claim path, summed its lock events, traced every OLAS transfer out of the Safe, decoded the veOLAS Deposit beneficiaries of the Safe's lock transactions, and read each account's current lock. Ingestion only, verdict human-set.

◆Established ● on-chain Corrects our earlier framing that only the 113.2M in buOLAS (about a quarter) was ever under enforced release: the veOLAS locks from the Sale contract were a larger, shorter-dated constraint. veOLAS locks are cliffs, not linear vesting, and carry voting power. No public label ties the 2-of-3 Safe to Valory, so the page keeps Valory's 10% as Olas's FAQ describes it.

OLAS is concentrated even after excluding infrastructure, though the two largest positions are protocol contracts rather than private wallets. 14,153 holders. The top 10 hold 75.39% raw and 55.67% excluding labelled infrastructure, with the two largest positions being protocol contracts: the governance Timelock at 14.14% and the buOLAS team-lock contract at 10.42%. Pulled the top holders with contract flags and public labels, classified each as holder/infra/insider, and computed raw and infrastructure-excluded top-10 concentration. Ingestion only, verdict human-set.

◆Established ● on-chain The two largest positions are the governance Timelock and the buOLAS team lock, and the Timelock is controlled by the same 5-of-9 Safe recorded in governance_authority. The infra-excluded figure rests partly on addresses Blockscout does not label, so a change in how many of those sit in the top ten calls for a trace by hand.

OLAS governance can act with no notice: the owning Timelock has a zero minimum delay, and a 5-of-9 Safe can propose and execute alongside the veOLAS Governor. Governance can act instantly. OLAS is owned by a TimelockController whose getMinDelay returns ZERO. Its proposers and executors are the current GovernorOLAS (0x060D...51E6, since 2026-06-15) and a 5-of-9 Safe, so the Safe can act without a vote. A timelock with no delay is a timelock in name: there is no window in which a holder can see a queued action and exit before it executes, which is the entire protection the pattern exists to give. Re-read 2026-09-24. Direct RPC read of the token contract's control surface: owner(), paused(), the two EIP-1967 proxy slots, Safe getThreshold/getOwners and TimelockController getMinDelay on whatever the owner resolves to, plus five public freeze getters. Shared implementation with npm run watch:control-surface, so the fact a human authors from and the fact the watcher re-reads are the same read. THE FREEZE PROBE IS A POSITIVE SIGNAL ONLY: FLOCK carries an admin-gated blacklist that every one of those getters reverts against, because its map is private, so a null there means no PUBLIC GETTER and never no blacklist. Scope is the token contract, not the project's off-chain services.

◆Established ● on-chain The delay is a parameter the Timelock's own admin can change, so a non-zero reading later would be a policy choice rather than a structural fix. The watcher tracks it either way.

Every contract in the OLAS token's ownership path, including the Timelock and both GovernorOLAS deployments, carries verified source. Confirmed at the contract layer: the OLAS token, the Timelock and both GovernorOLAS deployments all return verified source on Etherscan, which is what let us enumerate the governance roles above. Requested verified source metadata for each contract in the ownership path. Ingestion only, verdict human-set.

◆Established ● on-chain Verified source is what made the governance roles checkable at all. Scope is the contracts in the token's ownership path, not the wider agent framework repositories. Watched daily with the control surface, since a proxy upgrade could point at an unverified implementation.

The Olas component registry is small next to the service registry: a few hundred components against thousands of registered services. 328 components registered on the Ethereum L1 Component Registry NFT. Direct totalSupply read of the component registry contract, refreshed into meta.primary_data.

◆Established ● on-chain Components are the shared building blocks the architecture rests on. The ratio shows how thin the common code base is relative to the services registered against it.

Thousands of services are registered across the Olas registries on Ethereum and its L2s, counted cumulatively. 4,994 services registered cumulatively, summed across the ServiceRegistry NFT totalSupply on Ethereum L1 and the L2 registries. Sum of totalSupply on the L1 and L2 service registries, refreshed into meta.primary_data.

◆Established ● on-chain Cumulative, so it counts registrations rather than live services and can only rise. It is a registry footprint and says nothing about whether a registered service ever ran.

OLAS liquidity is thin and sits mostly in DEX pools, with MEXC the largest centralised venue. 22 tickers on 14 venues, re-read 2026-10-07 (25 on 2026-09-29). MEXC carries 46.2% of reported 24h volume, a Uniswap V2 pool on Ethereum 30.7% and Ourbit 17.9%; MEXC and Ourbit are the only centralised venues. The rest is Balancer pools across Gnosis, Base, Polygon and Arbitrum plus Orca, Aerodrome, Ubeswap and a Uniswap V2 pair on Robinhood Chain. Paginated the CoinGecko ticker set and summed 24h volume per venue; turnover from the trailing 7-day median of volume over market cap.

◆Established ◑ Cross-checked The venue shape is unusual: liquidity lives mostly in pools rather than exchange books, and a single CEX carries over half of reported volume. Carried as a concentration flag.

Olas's on-chain registries hold over a hundred registered agent blueprints and thousands of registered services, counted cumulatively. On-chain Olas registries (as of 2026-06-25): 115 registered agent blueprints, 328 registered components (Ethereum L1), and ~4,624 cumulative registered services across Ethereum/Gnosis/Base/Polygon/Optimism Sum of totalSupply on the registry NFTs. These are CUMULATIVE registered counts (one NFT per registration, never burned), an upper bound, NOT the marketed daily-active-agent population. Ingestion via refresh:primary, verdict human-set.

◆Established ● on-chain Cumulative registrations are an upper bound on participation and never fall, so they say nothing about daily activity. The daily-active-agent and transaction counters on olas.network measure a different population that these registries cannot test.

Ten addresses hold 81.63% of native VANA, and nine of them carry no explorer label. Native VANA top-10 is 81.66% of the 113,371,835 coin supply, and 81.63% survives the infrastructure exclusion because only one contract appears in the top ten, at 2.15%. The largest address holds 29.55% and the second 13.74%. 9 of the ten carry no label on the Vana explorer, so the insider share of 0% records absence of attribution rather than a measurement. Top-10 concentration in NATIVE VANA. Blockscout's token-holders endpoint covers ERC-20s only and cannot see a chain's own gas coin at all, so this reads the rich list at /api/v2/addresses and takes its denominator from the explorer's coin supply (113,371,834.86 VANA on 2026-09-06). Classification is the shared coldstart.sources.holders rule set, so contracts are neutral infrastructure and labelled insiders stay counted.

◆Established ● on-chain Measures the position on-chain, which is separate from the allocation plan. Nine of the ten are unattributed, so the composition behind the total is not settled. Whether the 8/15 distribution score can lead with the allocation plan at this concentration is flagged for the October review.

Vana's core contracts, SDK and personal-server code are public under permissive licences, but most of its repositories carry no licence file. Re-read 2026-09-18 through the GitHub API: 74 non-fork public repositories in github.com/vana-com (101 public in total, 27 of them forks), 47 with no licence file and 44 not pushed in 180 days. The core stack is permissively licensed and active: vana-smart-contracts is Apache-2.0, personal-server is MIT and vana-sdk is ISC, all pushed on the assessment date. Direct GitHub org/repo metadata reads. Ingestion only, verdict human-set.

◆Established ◑ Cross-checked Counted on a non-fork basis, because counting forks credits a project for code it did not write.

Gas paid on the Vana L1 is small, and Vana publishes no protocol-revenue figure. $2,348 in L1 gas fees over the trailing year, against $24,557 cumulative since launch and $145 in the trailing 30 days. DeFiLlama summary/fees/vana (dailyFees), refreshed into meta.primary_data.

◆Established ◑ Cross-checked DeFiLlama's vana entry is a chain module, so it reports validator-bound L1 gas, not a take rate on data-DAO activity. Banded on the trailing year, which can fall, rather than the all-time total, which only rises.

VANA trades on Binance and other large centralised venues, but Coinbase and Kraken do not list it. Binance, Bybit, KuCoin, Gate, Bitget, MEXC and CoinEx all quote VANA, across 30 tickers on 25 venues, with Upbit top at 25.5% of volume and Binance second. Coinbase returns NotFound for VANA-USD while the same call resolves IO-USD as online, and Kraken does not list it. Turnover is a 4.95% 7-day median. Paginated the CoinGecko ticker set, summed 24h volume per venue, then queried Coinbase directly for the named venue that did not appear rather than reading an aggregator gap as a delisting.

◆Established ◑ Cross-checked The venue that did not appear in the aggregator set was queried at the venue itself, so a gap is not read as a delisting.

Gas fees on the ZetaChain L1 are negligible, and neither ZetaChain nor Anuma publishes a revenue figure. $50.01 gas fees in 30 days, re-read 2026-10-07 ($32.26 on 2026-09-29, $13.57 on 2026-09-21), against $363,304 all-time (DeFiLlama chain adapter). GET of the DeFiLlama chain fees overview and chains list.

◆Established ◑ Cross-checked L1 gas is not protocol revenue. The 30-day figure is banded because the all-time total only rises.

Ten unattributed accounts hold 39.08% of issued PEAQ, one of them 19.01%, and none of the ten is a pallet account. Every account on the chain was enumerated rather than sampled: 3,393,045 System.Account entries against Balances.TotalIssuance of 4,446,080,550 PEAQ. Ten accounts hold 39.08% of issuance and one of them holds 19.01% (845,247,094 PEAQ) on its own; the next largest is 3.27%. Not one of the top ten is a pallet account, checked by the raw b"modl" prefix that identifies every pallet account whatever its id, so none of this is protocol plumbing. The largest positions are staked. The 19.01% account carries a Balances.Locks entry with lock id "peaqstak" over 249,555,360 PEAQ, and the 2.30% account carries one over 101,000,000 PEAQ of its 102,075,124 total. peaq's own explorer classifies both as Role: Nominator and shows 679 and 330 reward events respectively, so these are not dormant positions. Attribution to a named entity is still not possible: On-chain Identity reads Not Set, the Subscan profile is Hidden, and no label exists for any of the ten. Balances are free + reserved. A delegation on parachain-staking is locked in place rather than transferred, so a staker's full position is counted and this is ownership, not the liquid subset an Akash-style Cosmos bank read returns; the two are not comparable. Pallet accounts are identified by the b"modl" address prefix rather than by guessing pallet ids one at a time. Storage keys are built from a twox128 implementation checked against two prefixes already hardcoded in our own indexer (System.Account and Balances.TotalIssuance) before any new key was trusted, and every storage item is probed for chain-wide existence before an absence on one account is read as a measurement: Vesting.Vesting, Balances.Locks, Balances.Reserves, ParachainStaking.DelegatorState, CandidatePool and AtStake all return keys, while Identity.IdentityOf and four other ParachainStaking items are empty chain-wide and prove nothing. Ingestion only, verdict human-set.

◆Established ● on-chain Score held at 5/15. Attribution is missing in both directions: a 19% holder could be an exchange as easily as a founder, and unattributed is not evidence of insider control. Balances.Locks is the read that settles staking status; one absent staking storage item does not.

Collator entry on peaq is open because the candidate pool is below the slot cap, and collators post a fraction of a percent of the stake behind them. Collator selection is currently uncontested: there are fewer candidates than slots. 33 active collators against 42 slots, so selection is uncontested, with collators posting 0.19% of the stake and the rest delegated. Consensus set read from each chain's own endpoint by scripts/refresh-validator-sets.ts: active validators, stake shares and the NAKAMOTO HALT COEFFICIENT, which is the fewest validators whose combined stake exceeds one third. That is the BFT liveness threshold on these chains and it is NOT the 51% control figure the two get confused for, which is why the output carries nakamoto_basis. Censorship needs the halt number rather than the set size: a chain with a hundred validators and a coefficient of five is five colluding parties away from stopping.

◆Established ● on-chain Anyone who applies as a collator gets a slot while the pool is unfilled, which is open entry but also means no competitive pressure and no spare capacity if operators leave. Almost all stake is delegated, so block producers have little capital of their own at risk.

peaq's parachain node is public on GitHub under the Apache-2.0 licence and is actively maintained. Re-read 2026-09-18 through the GitHub API: 43 non-fork public repositories in github.com/peaqnetwork (100 public in total, 57 of them forks), 8 with no licence file and 32 not pushed in 180 days. The core peaq-network-node (the Substrate/Polkadot parachain node) is Apache-2.0, 81 stars, not archived, pushed 2026-07-20. The chain node and pallets are public under a permissive licence. Direct GitHub org/repo metadata reads. Ingestion only, verdict human-set.

◆Established ◑ Cross-checked Counted on a non-fork basis, because counting forks credits a project for code it did not write.

Application billing on peaq, led by MachineX, has recovered from its trough but remains far below its peak month. MachineX, the busiest application on peaq, billed $23,854 over the trailing 30 days, re-read 2026-09-24 ($21,201 on 2026-09-18, and $5,103 when this was first authored), so application billing keeps recovering from its trough while remaining far below the peak month of $166,363 in October. Trailing-30d fees annualised, with the monthly series read to establish the trend. Kept separate from the chain-gas row on purpose: summing an application's revenue into a chain's fee line is the error that produced the retired combined figure.

◆Established ◑ Cross-checked Kept separate from the chain-fee row because the two figures have different owners. Re-anchored 2026-09-18; the move that crossed the band was a recovery.

Little DeFi capital sits on peaq: chain TVL is well under a million dollars. $897,443 total value locked across peaq chains, re-read 2026-09-29 ($785,427 on 2026-09-24, $677,752 on 2026-09-18). DeFiLlama chain-TVL sum, refreshed by scripts/refresh-primary.ts into meta.primary_data.

◆Established ◑ Cross-checked Measures DeFi capital on peaq and says nothing about DePIN device economics. It bounds any argument that on-chain financial activity could fund the network. Re-anchored 2026-09-18.

PEAQ trades on Kraken, KuCoin, Gate, MEXC, Bitget and Crypto.com with thin but evenly spread depth; Binance has no PEAQ spot pair. 29 tickers on 23 venues on CoinGecko, re-read 2026-10-07 (40 on 2026-09-29, 47 on 2026-09-24). Kraken, KuCoin, Gate, MEXC, Bitget and Crypto.com Exchange all trade PEAQ, Binance's exchangeInfo returns PEAQUSDT as an invalid symbol, and depth is thin but evenly spread, the largest venue (Gate) at 12.6% of 24h volume. Paginated the CoinGecko ticker set, summed 24h volume per venue, then enumerated every Binance spot symbol with PEAQ as the base asset. Binance Alpha and futures were not checked and neither would be a main-spot listing.

◆Established ◑ Cross-checked No venue dominates the flow, so the book is thin but not concentrated. Binance Alpha and futures were not checked, and neither would be a main-spot listing.

IO's ten largest token accounts hold most of the supply, and most of that sits project-side in foundation multisigs and the vesting apparatus. The top 10 IO token accounts hold 58.82% of the 798,213,540 total supply, and all 10 owners are now labelled: 30.46% in two Squads multisig vaults Solscan names 'ocean foundation', 13.71% in the project's distribution and vesting apparatus, and 14.64% in exchange custody (2026-08-14). Read the 10 largest token accounts, resolved each to its owner, then classified every owner three ways: ed25519 curve membership (off-curve means no private key, so a program-derived address rather than a wallet), the programs its transaction history touches (SQDS4ep… identifies a Squads multisig vault, magnaSHy… the Magna vesting programme), and the source of its inbound IO traced one hop up. Explorer labels are attributed to Solscan, never inferred from an address prefix. Ingestion only, verdict human-set.

◆Established ● on-chain Exchange custody (Binance and Coinbase-associated accounts, per Solscan) accounts for 14.64%; stripping it out leaves 44.17% of total supply project-held. The raw top-ten share is re-measured by our refresher; the labelled split was traced by hand on Solscan on 2026-08-14 and is not re-measured, so a change in composition behind an unchanged total would not route.

The IO mint has no freeze authority and no mint authority, so nobody can freeze an IO account or issue more IO. The IO SPL mint has freezeAuthority = NULL: no authority can freeze an IO token account, so there is no token-layer censorship lever. Combined with the null mint authority the token is immutable at the SPL layer. Direct keyless Solana RPC read of the mint account's authority fields. Ingestion only, verdict human-set.

◆Established ● on-chain Speaks to the token only. Whether io.net can refuse or remove a workload or supplier on its own orchestration platform is an off-chain question and remains unassessed.

io.net's reported client burns reconcile with the chain epoch by epoch, which puts a floor under its reported IO purchases. Floored on chain, with the purchase figure itself remaining io.net's. On 2026-09-19 the newest 294 settled epochs, five weeks from 12 August to 18 September, were reconciled one by one: each epoch's reported burn was matched against the SPL burn on the IO mint in that epoch's on-chain settlement transaction, and 294 of 294 match to six decimals with 0 mismatches (688,598.285 IO on chain against 688,598.285 reported). The cumulative burn is separately verified against the mint's supply reduction. io.net reports 5,802,542 IO purchased and 2,146,039 burned, a 37.0% burn share; with the burn verified on chain and the share published, purchases cannot be below 5,802,542 IO on io.net's own accounting. What is NOT on chain is the purchases themselves, which settle in fiat and USDC as well as IO, so the purchase total stays a reported figure with a verified floor rather than a measured one. Walked io.net's own epoch list, resolved each epoch_signature as an account pubkey, and found the burnChecked instruction on the IO mint among that account's settlement transactions. Reconciled per epoch rather than in aggregate, so a single epoch whose reported burn diverged from the chain would surface as a mismatch instead of vanishing into a total.

◆Established ● on-chain The purchases themselves are not on chain: io.net's docs price compute in USDC or IO, so the purchase total stays io.net's figure with a checked floor. Every epoch's burn is on chain and reconciles.

IO trades actively across dozens of venues, including Binance and Coinbase, with high daily turnover for its market cap. Binance quotes IO at $935,280, second by volume across 73 tickers on 59 venues totalling $7.58M, and turnover is a 13.86% 7-day median: roughly a seventh of the market cap changes hands daily. The leader is Aivora Exchange at 13.5%, with Coinbase also quoting it. Paginated the CoinGecko ticker set and summed 24h volume per venue; turnover from the trailing 7-day median of volume over market cap.

◆Established ◑ Cross-checked Aivora is a little-known venue and leads volume on other tokens too, so its share is read with caution. The drawdown from ATH is a separate question and is not graded here.

The active validator set is stake-weighted and concentrated: seven entities hold a third of active escrow. 264 registered entities, of which 75 are in the active set. Active escrow concentration: top 1 at 11.1%, top 5 at 30.1%, top 10 at 42.7%, top 20 at 64.6%. Seven entities reach a third of active escrow and fourteen reach half. Named entities lead the set (Colossus, BinanceStaking, Mars Staking), so the largest holders are identifiable rather than anonymous. Pulled all 264 registered validators, filtered to the active set, summed active_balance escrow and computed top-N shares plus the entity counts crossing one third and one half. First-party indexer, hence api grade. Verdict human-set.

◆Established ◑ Cross-checked The largest holders are named entities. A third of stake is the Byzantine threshold, which is why the seven-entity figure matters on a confidentiality chain.

One unlabelled account holds a quarter of all ROSE, and its transaction pattern marks it as a custodial reserve whose owner is unnamed. Ten accounts hold 57.6% of the 10B supply, 54.53% after excluding the Common Pool (5.07%). One unlabelled account holds 25.52% of all ROSE (2,552,104,444 ROSE, entirely liquid, zero escrow and zero delegations), the second 6.71% and the third 5.21%, both mostly delegated. Nine of the ten carry no label anywhere Oasis publishes, so the insider share of 0% records absence of attribution rather than a measurement. The largest account has made 303 transactions since April 2021, and 81 of them face a single counterparty, oasis1qzwfdgpt3p6dd2mk6207qhf9qrxuvfn9eunurdgu, across 2,536,644,426 ROSE. That counterparty has made 270,818 transactions at nonce 173,465 while retaining only 39,236,632 ROSE, and its live feed is continuous two-way transfers with many distinct counterparties in unrounded amounts. An account that has sent over 170,000 transactions to many parties while holding almost nothing is withdrawal infrastructure, so the 25.52% holder is a reserve funding a hot wallet that serves many beneficiaries. Which custodian is not established: neither address carries a name on the Foundation's own explorer or in its Nexus indexer. Rich list from the Foundation-run indexer against its own total supply; protocol accounts from oasis-core's reserved-address list. Function is then established from transaction structure rather than from a resembling balance: counterparties are counted, and the distinguishing evidence is the dominant counterparty's own nonce and transaction count, which measure how many parties it pays. That is a measurement of many beneficiaries, not an inference from a shape, and the distinction matters because attributing a large holding on a pattern that merely looked like custody is the error the Flux row records. Naming the custodian would need a label source and none exists. Ingestion only, verdict human-set.

◆Established ◑ Cross-checked Counted in the rated figure on purpose. Excluding the custodial reserve would put the infra-excluded top ten near 29%, and that reclassification rests on behaviour rather than on a name, so it is a review decision. Unlabelled is unattributed in both directions, so the exchange is not named.

Eight validators together hold more than a third of stake, enough to halt the chain. Re-read 2026-10-05 from the chain's own endpoint: 74 active validators, halt coefficient 8 (the fewest whose combined stake exceeds a third), top-10 share 40.99%. It was 6 on 2026-09-18. Consensus set read from each chain's own endpoint by scripts/refresh-validator-sets.ts: active validators, stake shares and the NAKAMOTO HALT COEFFICIENT, which is the fewest validators whose combined stake exceeds one third. That is the BFT liveness threshold on these chains and it is NOT the 51% control figure the two get confused for, which is why the output carries nakamoto_basis. Censorship needs the halt number rather than the set size: a chain with a hundred validators and a coefficient of five is five colluding parties away from stopping.

◆Established ● on-chain The halt threshold and the stake held by the largest handful bind censorship resistance here, more than the size of the set. TEE and side-channel questions are a separate layer and are not measured by this row.

ROSE is listed on Binance, Coinbase, KuCoin and Bybit among 25 venues, with moderate turnover. ROSE quotes across 30 tickers on 25 venues, and Coinbase returns ROSE-USD as online. Binance leads at 40.3% of volume, with BTCC, KuCoin, WhiteBIT, Coinbase and Bybit behind it. Turnover is a 5.65% 7-day median. Paginated the CoinGecko ticker set and summed 24h volume per venue, and confirmed the Coinbase listing at Coinbase.

◆Established ◑ Cross-checked The Coinbase listing was confirmed at Coinbase directly.

Most of ORA's supply is insider-controlled: two team Safes and a Safe-funded wallet hold about 77% of it. Ethereum ORA top-10 holders = 94.7% raw; infra-excluded (the LayerZero OFT adapter) still 81.4% of supply. 50.1% sits in two Gnosis Safe multisigs (0x8Eb8… 27.1% and 0x5e546… 23.0%, team, treasury or foundation); the funding trace then resolves the third-largest holder, a 26.3% EOA (0x8520…), as Safe-funded (from:SafeProxy), so it too is insider, lifting insider concentration to ~77%. The launch float may have been sold to the community, but the supply is ~77% insider-controlled (D0). sources/holders.py: top holders + is_contract + labels, classified holder/insider/infra/cex/burn; Safes counted as insider concentration, the OFT bridge adapter excluded as neutral infra. Concentration = raw balance / total_supply. Ethereum is ORA's canonical chain (OFT); a minority is bridged to spokes.

◆Established ● on-chain ORA's own ORA Coin page described the launch as a Decentralized AI Community Offering inspired by the DAICO idea; that describes how the float was sold, which this measurement does not contradict. The same page listed circulating supply at 36,666,666.63 of 333,333,333. The infra-excluded figure rests partly on addresses Blockscout does not label, so a change in how many of those sit in the top ten is a signal to re-trace by hand.

Nobody can freeze, seize or claw back an ORA balance: the token has no pause, blacklist or admin role and cannot be upgraded. At the token layer, the verified source contains no blacklist, no pause modifier, no Ownable and no AccessControl, and the deployed contract reverts on every corresponding call. Nobody can freeze, seize or claw back an ORA balance, and nobody can replace the logic that would allow it, because the contract is not upgradeable. This is a narrower finding than network-level censorship resistance: ORA's inference runs off-chain and is not covered by it. Same runtime probe as the supply verdict, read for the freeze question: absence of pause/blacklist in the verified source, corroborated by reverts on the deployed contract, plus zero EIP-1967 slots to rule out a future upgrade adding them. Scope stated deliberately: this covers the token, not the off-chain inference layer, per the verdict spec's scope-matching rule. Verdict human-set.

◆Established ● on-chain Token layer only. ORA's 9/15 on this dimension is reasoned from the opML AnyTrust model and off-chain inference, a separate scope. On the daily control-surface watch since 2026-09-18: a getter hit is a fact, while a miss is recorded as "no public getter answered", never as "no blacklist exists".

ORA's documentation site, docs.ora.io, has been deleted, and the protocol's docs survive only in Internet Archive captures. docs.ora.io returns NXDOMAIN. The Wayback CDX index shows docs.ora.io serving full content as late as 2026-04-22, and www.ora.io serving the old oracle homepage (digest JLT5REU75WO6CLXEQMG66PUAHTM7E4OJ) as late as 2026-05-05, with the replacement /funds page first captured 2026-06-07. So the documentation was retired alongside a product pivot in May-June 2026 rather than moved. What survives: the github.com/ora-io org, and the blog at paragraph.com/@orablog whose most recent post is dated 13 January 2025. foundation.ora.io, printed in ORA's own docs as the foundation's home, no longer resolves either. Queried three independent public resolvers for an A record and captured the response status, with a known-good hostname on the same domain as the control, then confirmed the failure in a browser. Separately enumerated the GitHub org and walked the blog's post index to establish what first-party publication survives. Verdict human-set.

◆Established ◑ Cross-checked Settled, so the re-check cadence is 180 days. Search engines still index docs.ora.io, so published links dead-end for readers. The apex ora.io is now dark too; that is recorded on the product_continuity row.

ORA's GitHub org publishes a Salus Security audit of its staking contract; no published audit covers the opML or oracle core. The repository ora-io/audit-report-staking contains ORA_staking-contract_audit_report_2024-07-13.pdf: a Salus Security review of ORA's staking contract at version v5, commit f4ddf14, engagement logged across 21 June to 13 July 2024. It reports 2 high-severity, 3 medium-severity and 6 low-severity issues, with 11 findings marked Resolved, 1 Mitigated and 1 Acknowledged. The audited source sits alongside it in ora-io/staking-contract-audit. The scope is the staking contract only: no published audit was found for the opML or OAO core, and the ORA token contract itself needs none, since it is an unmodified OpenZeppelin ERC20Permit/ERC20Burnable with no admin surface. Enumerated the org's repositories, downloaded the report and extracted its text to read the auditor name, date, contract version, commit and per-severity counts, rather than relying on the repository name. The named-firm-plus-public-report bar in the open-source rubric is met. Verdict human-set.

◆Established ◑ Cross-checked A named firm with a public, dated report clears the open-source rubric's audit bar. The score is held because the audit covers staking rather than the opML core, which carries the thesis. A project's GitHub org is a publication surface and is checked before any absence of audits is recorded.

ORA's supply cannot rise: the token has no mint function, no owner or role admin, and sits behind no proxy. The runtime checks are independent of the published source: both EIP-1967 slots read zero, so the token is not behind a proxy and its logic cannot be replaced; owner(), getOwner(), hasRole(), DEFAULT_ADMIN_ROLE(), MINTER_ROLE(), minter() and paused() all revert, so there is no owner, no role admin, no minter and no pause; and totalSupply() returns exactly 333,333,333000000000000000000, matching the TOTAL_SUPPLY constant. The verified source carries no mint function at all: ORACoin is ERC20Permit plus ERC20Burnable, and the only _mint calls are the four in the constructor, closed by assert(totalSupply() == TOTAL_SUPPLY). Supply can fall through burns and cannot rise. Probed the deployed contract for every standard control surface (ownership, role admin, minter, pause) and read both EIP-1967 slots directly, so the immutability finding rests on runtime behaviour rather than on reading the project's own source. The source was then read to confirm no mint path exists and to locate the constructor's four _mint calls. Verdict human-set.

◆Established ● on-chain The supply is beyond anyone's control while roughly 77% of it sits with insiders; those are different questions and both answers stand. ORA's published total supply is recorded on the total_supply row.

ORA trades only in a handful of thin DEX pools; no centralised exchange quotes it. Four tickers on four venues, all decentralised: Uniswap V3 on Ethereum at 83.5%, PancakeSwap V3 on BSC, Uniswap V3 on Base and a Raydium CLMM pool. No centralised venue quotes ORA. Total reported 24h volume across every venue is $152, so turnover reads a 0.047% 7-day median and a single ordinary order would move the price. Paginated the CoinGecko ticker set and summed 24h volume per venue; turnover from the trailing 7-day median of volume over market cap.

◆Established ◑ Cross-checked Any order of ordinary size carries heavy price impact at this depth.

Once the LayerZero bridge adapter and other infrastructure are excluded, GIZA's top ten holders hold a modest share, but the holder base is thin. The raw top-10 number is alarming and the real one is not. Top-10 holders control 97.72% of supply, but 88.53% of that is a single LayerZero OFT adapter (the bridge lock holding supply that circulates on other chains) and the labelled infrastructure share is 88.89%. Excluding infrastructure, the top 10 hold 8.92% across 2,186 holders, with the largest non-infra holder being the team Safe at 8.49%. Pulled the top holders with contract flags and public labels, classified each as holder/infra/insider, and computed both the raw and infrastructure-excluded top-10 concentration. Ingestion only, verdict human-set.

◆Established ● on-chain Raw top-ten concentration is dominated by a single LayerZero OFT adapter, the bridge lock for supply circulating on other chains. The thin holder base is the more meaningful weakness. The infrastructure-excluded figure rests partly on addresses Blockscout does not label, so a change in how many of those sit in the top ten is a signal to re-trace by hand.

The GIZA token has no pause function and no proxy, so transfers cannot be halted and its logic cannot be replaced; a 2-of-4 Safe owns it, and its mint is capped at the supply already issued. GIZA is owned by a 2-of-4 Safe. Its only privileged function, mint, is capped in code at the 1,000,000,000 maxSupply, which is fully issued. There is no pause function and no proxy, so the logic cannot be replaced and transfers cannot be halted. Direct RPC read of the token contract's control surface: owner(), paused(), the two EIP-1967 proxy slots, Safe getThreshold/getOwners and TimelockController getMinDelay on whatever the owner resolves to, plus five public freeze getters. Shared implementation with npm run watch:control-surface, so the fact a human authors from and the fact the watcher re-reads are the same read. THE FREEZE PROBE IS A POSITIVE SIGNAL ONLY: FLOCK carries an admin-gated blacklist that every one of those getters reverts against, because its map is private, so a null there means no PUBLIC GETTER and never no blacklist. Scope is the token contract, not the project's off-chain services.

◆Established ● on-chain The dimension's custody point, that smart-account assets stay with users if Giza pauses its hosted optimiser, is outside this token-contract read.

Giza's code is public on GitHub, but development has stalled: no repository has been pushed since 23 March 2026. Re-read 2026-09-18 through the GitHub API: 28 non-fork public repositories in github.com/gizatechxyz (32 public in total, 4 of them forks), 19 carrying no licence file and 27 not pushed in 180 days. The newest push is giza-hub on 2026-03-23; LuminAIR (the proving library, 57 stars) and giza-token last moved in September 2025. The token contract source is verified on Etherscan. Enumerated the org's repositories by last-push date. Ingestion only, verdict human-set.

◆Established ◑ Cross-checked Counted on a non-fork basis, because counting forks credits a project for code it did not write. Development stopped around the same March 2026 date the agent dashboard froze, which corroborates the retirement of ARMA and Pulse.

GIZA staking is live on Base, and the staked balance is shrinking. 38,410,539 GIZA staked in the Staker contract (Base), re-read 2026-09-18. balanceOf via eth_call, scaled 18

◆Established ● on-chain Directly measurable on-chain. Staked GIZA fell from 46,298,698 at authoring to 38,410,539 on 2026-09-18, about 17%, alongside the fee revenue that rewards it. Giza's docs say operator staking details are still to come. Separate from agent AUA (aua).

GIZA liquidity is very thin: no centralised listing remains and CoinGecko no longer tracks a market for it. 2 tickers on CoinGecko, re-read 2026-10-07 (0 on 2026-09-29, 1 on 2026-09-24): an Aerodrome Slipstream pool and a Uniswap V3 pool, both on Base, with negligible 24h volume. No centralised exchange quotes GIZA. Paginated the CoinGecko ticker set, summed 24h volume per venue, then queried Gate and MEXC directly. Five simultaneous absences on a token trading $164 a day is a delisting pattern rather than an aggregator gap, and two venues confirmed it explicitly.

◆Established ◑ Cross-checked Gate and MEXC confirmed the delisting directly. Re-read 2026-09-21: unchanged at 2 tickers on 0.149% turnover. The liquidity dimension score is flagged for a full re-read at the October review. Re-read 2026-09-29: CoinGecko dropped the last ticker; the Aerodrome pools remain on DexScreener.

Ten unattributed addresses hold close to half of ALLO supply, and no vesting account appears among them. Ten addresses hold 46.51% of the 788,127,838 ALLO supply, the largest 11.26%, across 13,271 holders in total. None of the ten is a module account and none is a vesting account: the whole top ten is unattributed, so the 0% insider share records that nobody has traced them. The infra-excluded figure equals the raw one because no neutral infrastructure reaches the top ten, and staking is small: the bonded pool holds about 1.3% of supply and the chain's own staking module about 0.7%, both far below the tenth-largest holder. Every ALLO holder enumerated through the chain's own bank module (cosmos/bank/v1beta1/denom_owners/uallo, keyless, three pages of 5,000), sorted by balance, against the bank supply. The top ten is exact rather than a best-effort from a rich list, because the whole owner set is read. Cosmos publishes no explorer labels, so classification comes from the chain: module accounts are derived from cosmos/auth/v1beta1/module_accounts and excluded as neutral infrastructure, and each unattributed top address is asked of the auth module directly so a vesting account counts as insider. Shared with the Warden reader in coldstart.sources.holders.

◆Established ● on-chain The top-ten total sits close to the stated insider allocation, and that proximity is not evidence: we do not attribute these addresses to early backers on similar percentages. No chain-level vesting account enforces a lock among the ten, though the allocation could sit with a custodian or a custom module; neither was established. A labelled trace is due before the score moves.

A handful of validators together hold more than a third of stake, enough to halt the chain, and little of the supply is bonded. 17 active validators, and the Nakamoto halt coefficient is 6: that many colluding validators exceed one third of stake and can stop the chain. Top ten hold 62.74% of stake, and 1.34% of supply is bonded. Consensus set read from each chain's own endpoint by scripts/refresh-validator-sets.ts: active validators, stake shares and the NAKAMOTO HALT COEFFICIENT, which is the fewest validators whose combined stake exceeds one third. That is the BFT liveness threshold on these chains and it is NOT the 51% control figure the two get confused for, which is why the output carries nakamoto_basis. Censorship needs the halt number rather than the set size: a chain with a hundred validators and a coefficient of five is five colluding parties away from stopping.

◆Established ● on-chain Coordinating a halt needs far fewer parties than the set size suggests, and the stake defending the chain is small as well as concentrated. Censorship Resistance 9/15 rests on inherited Cosmos properties and is flagged for the October review.

Allora's core chain is public under Apache-2.0, though about half of the org's own repositories carry no licence file. Re-read 2026-09-18 through the GitHub API: 11 non-fork public repositories in github.com/allora-network (14 public in total, 3 of them forks), 6 carrying no licence file and 2 not pushed in 180 days, newest push docs on 2026-09-18. The core chain allora-network/allora-chain is Apache-2.0 and not archived. Direct GitHub org/repo metadata reads (public_repos, licence, stars, pushed_at). Ingestion only, verdict human-set.

◆Established ◑ Cross-checked Counted on a non-fork basis, because counting forks credits a project for code it did not write.

Issued ALLO is well short of the 1B maximum, and the first cliff tranche, due in November 2026, is worth roughly two-thirds of the current float. About 788.1M of the 1B max has been issued, so circulating is about a quarter of max supply and closer to a third of what exists. The first cliff tranche is worth roughly two-thirds of the current float. Total issued ALLO from the chain's own bank module (cosmos/bank/v1beta1/supply/by_denom?denom=uallo), keyless, against the 1B max the tokenomics declares. Circulating and the FDV/MCap ratio come from the batched CoinGecko read already stored in src/data/market-data.json, so no absolute market figure is restated in prose: the ratio and the supply percentages are what the verdict turns on.

◆Established ● on-chain A percentage of supply needs its denominator: with a 1B max and 788M issued, the same event reads as 24% or 32% of the float. Supply Dynamics 8/20 held; the November cliff belongs in the October review.

ALLO trades on a broad set of centralised venues, and its turnover sits mid-pack among the projects we measure. ALLO trades on Binance, Coinbase, Kraken, OKX, KuCoin, Gate.io and MEXC, across 39 tickers on 29 venues, with Binance carrying 28.2% of reported 24h volume and no second venue above 22%. Turnover is a 7.19% 7-day median, mid-pack in our own corpus: Virtuals reads 12.05% and Aethir 7.51% on the same measure and the same day, before counting tokens whose higher ratios are one-venue artefacts (Nillion 42.78% on LBank, IoTeX 36.24% on BitDelta). Paginated the CoinGecko ticker set and summed 24h volume per venue, then compared the turnover median against every other project carrying the same metric on the same day.

◆Established ◑ Cross-checked Listing breadth is strong for a token this size.

FLUX custody is concentrated: ten addresses hold most of the supply, and one unattributed address holds over a third of it. Ten addresses hold 59.43% of the 419,769,138 FLUX the explorer reports as supply, and one P2SH address (t3ThbWogDoAjGuS6DEnmN1GWJBRbVjSUK4T) holds 38.12% of it on its own. Without that one address the top ten is 21.32%, and without the top two it is 11.91%. The largest address has 23 transactions in its life, has received 277.0M FLUX and sent 117.0M, and moves in round 5M and 10M chunks to a single counterparty, which is the behaviour of a vault rather than a holder. The second (39.5M, 9.41%) has four transactions. Neither is attributed anywhere Flux publishes. Several of the rest carry large blocks_mined counts and read as mining pools. The listed thousand addresses hold 92.14% of supply between them. Rich list from the project-run insight explorer against the same explorer's supply figure. No label map exists, so `known` is empty and nothing is classified as infra; the raw and ex-infra figures are therefore identical by construction, and that is a statement about attribution rather than about the chain. The largest address was then traced by transaction history, which narrowed what it behaves like without establishing what it is. Ingestion only, verdict human-set.

◆Established ◑ Cross-checked All ten of the top addresses are unattributed, so the figure is an upper bound on concentration. Flux's FluxNode documentation describes its parallel assets as mining rewards rather than a 1:1 swap of locked FLUX, so the largest address cannot be classed as bridge custody. Its history reads as a distribution wallet: 23 lifetime transactions, and a sole large counterparty that has passed 723.4M FLUX through 406 transactions.

No single operator or host can stop the Flux network, but 38 payout addresses run half the nodes and two Hetzner blocks hold 727. No single point of failure, though concentrated on the operator and hosting axes (index read 2026-09-28, block 2,991,152). Operators: the largest single payment address runs 415 nodes (6.3% of the network), the top 10 run 28.9%, and 38 addresses run half, while 361 addresses run exactly one node each. Hosting: the 6,570 nodes sit in 564 distinct /16 IP blocks, but the top 10 blocks hold 34.8% of them and 25 blocks hold half. RIPE RDAP registers the single largest block to GHOSTnet GmbH (5.230.0.0, 499 nodes), the two Hetzner Online blocks (65.109.0.0 and 65.108.0.0) to 727 nodes between them, and 62.107.0.0 to Norlys Telco A/S, a Danish ISP group (407 nodes). Block production is the healthiest axis: 234 distinct producers over a 2,875-block window, the largest at 11.58%, and 12 producers needed to reach half the blocks. Concentration computed from the node list by payment address and by IP /16. Network operators were resolved through RIPE RDAP rather than inferred from the address prefix, and are attributed to the registry record rather than asserted. Block-producer concentration read from the explorer's 2,880-block window. The RDAP leg is genuinely independent of Flux; the node list is not. Verdict human-set.

◆Established ● on-chain The operator and hosting index that rates infrastructure decentralisation also answers the takedown question. An actor pressuring hosting providers has far fewer parties to approach than the country count suggests. RIPE RDAP attribution for the largest blocks refreshed with the index on 2026-09-28; GHOSTnet replaced Norlys as the single largest block.

Flux's core node code is public under AGPL-3.0, but a large share of its own repositories carry no licence file. Read 2026-09-18 through the GitHub API: 110 non-fork public repositories in github.com/RunOnFlux (189 public in total, 79 of them forks), 79 carrying no licence file and 60 not pushed in 180 days, newest push fluxhashes on 2026-09-17. The core flux node repo is AGPL-3.0 and not archived. Direct GitHub org/repo metadata reads. Ingestion only, verdict human-set.

◆Established ◑ Cross-checked Counted on a non-fork basis, because counting forks credits a project for code it did not write.

Most of the 560M FLUX cap is already mined, and daily block rewards keep adding to circulating supply. The explorer's supply statistics put mined supply at 430,232,792.5 FLUX on 2026-09-20, 76.8% of the 560M cap, growing by about 40K a day. The explorer's separate total-supply endpoint reads 419,744,862 on the same day, 2.4% below the statistics series, and neither endpoint documents its basis, so the two are not reconciled here. The 560M cap is a fork parameter that no endpoint exposes; it is carried as the project's statement. published-json json_path 0.sum, cross-read against the explorer's total-supply endpoint. Chain-derived figures on a project-run explorer, hence api grade. Ingestion only, verdict human-set.

◆Established ◑ Cross-checked Wired to the explorer's supply statistics, which re-read mined supply each refresh.

FLUX lists on Binance, Kraken, KuCoin, Gate and Crypto.com, but more than half of reported volume sits on one venue, GroveX. Binance, Kraken, KuCoin, Gate and Crypto.com Exchange all list FLUX, among 42 tickers on 34 venues: Binance $94,248, Kraken $41,084. GroveX alone carries $1,666,271 of $3.05M reported 24h volume, 54.7%, roughly eighteen times Binance's share of the same book. Turnover reads an 18.89% 7-day median on that basis and roughly 8.6% without GroveX. Paginated the CoinGecko ticker set and summed 24h volume per venue, then recomputed turnover with the dominant venue excluded to see what the headline rests on.

◆Established ◑ Cross-checked With more than half the flow on one venue, the turnover ratio is not a depth measure here. Carried as a concentration flag rather than a mark against the score.

FluxCloud's application registry is broad, with over a thousand distinct owners, though one owner holds a large share of the specifications. The global application registry holds 1,872 specifications, of which 1,823 are unexpired at block 2,991,152 (read 2026-09-28; 1,018 of 1,074 at block 2,865,303 in August), requesting 8,909 instances across 1,283 distinct application owners. The largest single owner accounts for 201 specifications, about 11% of the unexpired set. Pulled every global application specification and filtered to those whose height plus expire exceeds the current block height, so expired registrations do not inflate the count. Counted requested instances and distinct owners. First-party registry, hence api grade. Verdict human-set.

◆Established ◑ Cross-checked These are unexpired registrations requesting instances, not confirmed running containers, so the count measures demand placed on the network rather than delivered uptime.

Across the Cardano and Ethereum legs, ten holders hold about 38% of NTX, or about 18% once exchange and infrastructure positions are excluded. Holders merged across the Cardano and Ethereum legs, rated against all three leg supplies (1,408,013,812 NTX: 966,647,334 Ethereum, 429,366,478 Cardano, 12,000,000 BSC). Ten holders hold 38.09%, falling to 18.20% once exchange and neutral-infrastructure positions come out. The single largest is a Cardano stake key at 7.30% (103.3M NTX), unattributed. The infra share is 22.89% and is dominated by two Ethereum venues, Uniswap V4's PoolManager at 6.76% and a Uniswap V2 pair at 2.58%, which is where the attacker's newly minted NTX was sold on 19 and 20 September. Only one top-ten position is attributable to an insider: the SingularityNET Treasury Safe at 2.13%. Cardano holds 7,172 holders by stake key. Four of the top ten carry no label anywhere either chain publishes. Cardano and Ethereum are different address spaces, so merging them cannot double-count one holder, which is what made this token rateable where a two-EVM-leg token still is not. Cardano balances are aggregated by STAKE KEY rather than payment address, because one wallet spreads a balance across many addresses and counting them separately would understate concentration. Ethereum rows are classified by the same labelled-holder rules the rest of the corpus uses. The BSC leg (0.85% of supply) has no keyless holder source: its supply is in the denominator and its holders are not read, so concentration is understated by at most that. Ingestion only, verdict human-set.

◆Established ● on-chain Read this with the date attached: 408.5M of the Ethereum leg is attacker-minted NTX from 19 September, and the DEX positions in the infra share are where it was sold, so the figure describes an incident in progress. Only 2.13% of supply sits at a top-ten address attributed to an insider. Score held at 6/15. Re-read once the Ethereum leg is unpaused and the supply question is resolved.

The compromised deployer key still controls upgrades of the BSC NTX token, so the BSC leg was not insulated from the September 2026 compromise. The BSC NTX token 0x5c4bcc4dbaeabc7659f6435bce4e659314ebad87 is an EIP-1967 proxy. Its admin slot holds ProxyAdmin 0xf0d33beda4d734c72684b5f9abbebf715d0a7935, whose owner() is 0x863f13e5b505f1eb17803b94ec9d3daf80092165, the deployer EOA that minted 408.5M NTX on Ethereum on 2026-09-19. The ProxyAdmin exposes upgrade and upgradeAndCall, so that key can replace the token's implementation, currently 0xd760b2a63ccb0d7049b75abd86e0e1fd204f1d3b, which has no mint function of its own. The token's own owner() moved by 3 October 2026 from the EOA 0x78a60de4fbf1f1c2daf8c94b5e40f877032cef00 (the address that took admin of the Ethereum leg after the mint) to 0x8714330e007ccc950ca2e651035b4893d0806ed4, a Safe v1.5.0 with a 2-of-3 threshold and the same three signers as the Ethereum admin Safe. The ProxyAdmin's owner was unchanged on a 2026-10-07 re-read. BSC totalSupply() reads 12,000,000 NTX. Direct RPC reads of proxy storage and ownership. The ProxyAdmin sits at the same address as the Ethereum NuNetToken, so the two must not be confused.

◆Established ● on-chain Corrects our earlier framing that the BSC leg was untouched by the compromise. It was not paused, but the key that minted on Ethereum can still upgrade it. Censorship resistance score unchanged pending review.

The Ethereum NTX contract is paused, freezing every transfer on that leg; admin and pause control have sat with a 2-of-3 Safe since 29 September 2026. The pause is no longer hypothetical: paused() returns true. After the unauthorised mint the roles were rotated through the compromised admin itself. Block 26017576: DEFAULT_ADMIN_ROLE granted to a new EOA 0x78a60de4fbf1f1c2daf8c94b5e40f877032cef00. Block 26017604 (08:43:47 UTC): DEFAULT_ADMIN_ROLE revoked from the original deployer EOA 0x863f13e5b505f1eb17803b94ec9d3daf80092165. Block 26017632: PAUSER_ROLE granted to the new EOA. Block 26017641 (08:51:11 UTC): pause() called, freezing every NTX transfer on Ethereum for every holder, with no owner carve-out. That is 12 hours and 1 minute after the mint. The freeze is leg-local: the Cardano native asset and the BSC leg were not paused. The BSC leg is not out of the compromised key's reach, though: its upgrade authority still sits with that key (see bsc_upgrade_authority). Control then passed to a multisig. Block 26074970 (28 September 2026): the new EOA granted DEFAULT_ADMIN_ROLE to 0x2f17d7cc940124f539952641360b92f003c96079, a Safe v1.5.0 with a 2-of-3 threshold over three EOA signers. Blocks 26082789 and 26082798 (29 September): the EOA revoked its own admin and moved PAUSER_ROLE to the Safe. Re-read 2026-10-07: the Safe is the only DEFAULT_ADMIN_ROLE member and holds PAUSER_ROLE, the EOA holds neither, MINTER_ROLE's only member is the contract 0x6c0d706c75b559549938c0b1de863cf7f042d1cf, and paused() still returns true. Manual read of the verified contract (pause path, no blacklist/fee logic) + live pause/role state + owner classification. Ingestion only, verdict human-set.

◆Established ● on-chain Severity raised info -> high and censorship_resistance cut 9/15 -> 7/15. The score keeps most of its credit because the pause was leg-local, and loses two points because a single un-renounced EOA froze the largest leg for every holder and handed control to another bare EOA. On 28-29 September 2026 admin and pause moved to a 2-of-3 Safe; the censorship_resistance cut is held pending review.

NTX trades thinly on MEXC and a handful of DEX pools, with no tier-1 exchange listing. 4 tickers on 3 venues on CoinGecko, re-read 2026-10-07 (5 on 2026-09-29, 9 on 2026-09-24): MEXC, Minswap on Cardano and PancakeSwap V2, with combined 24h volume in the hundreds of dollars. WingRiders and the Ethereum Uniswap pools no longer appear in the set, and no tier-1 CEX quotes NTX. Paginated the CoinGecko ticker set and summed 24h volume per venue; turnover from the trailing 7-day median of volume over market cap.

◆Established ◑ Cross-checked A thin market led by MEXC and DEX pools.

Compute providers are registered and active on the on-chain provider registry. The registry enumerates to 61 active providers, read from getActiveProviders on the LumerinDiamond, re-read 2026-09-18. Direct registry enumeration on both deployments, summed and split per chain. Replaces a MorScan status figure that had no independent leg; MorScan is a MOR-earning builder subnet and cannot check Morpheus on its own account.

◆Established ● on-chain Registered is not the same as serving: our own SessionClosed index shows 29 distinct providers that have closed a session with a receipt, a floor on the serving set.

One 5-of-9 Gnosis Safe holds owner and upgrade authority over the core deposit, staking and inference contracts, with no timelock. One 5-of-9 Gnosis Safe (0x1FE04BC1...) holds owner/upgrade authority across the core contracts on both chains: Base Builders staking (0x42BB446e, UUPS, no timelock), Base LumerinDiamond inference, and the Ethereum-L1 Distribution contract where capital deposits sit (0x47176B2A..., upgradeable proxy, no timelock). The MOR OFT token is on a separate Safe (0xf3ef0016...). Traced owner/upgrade authority across the core contracts on both chains to the 5-of-9 Safe; confirmed no timelock.

◆Established ● on-chain Governance rubric stage G0: a single 5-of-9 Safe, no timelock, over the deposit, staking and inference contracts. A timelock would reach G1; binding MRC or Snapshot votes executed through it, with the Safe cut to emergency scope, would reach G2. Verified across Base core and the L1 deposit contract; the Arbitrum reward contract and signer independence are the remaining refinements. Watched from 2026-09-17: the Safe is probed on both chains, alongside the L1 Distribution contract and the LumerinDiamond it owns, so a change in who can upgrade them routes to a human. Morpheus describes its governance as Atomic Governance, with final judgement resting with repository owners; that statement is recorded in the atomic_governance row.

The Morpheus AI Snapshot space (morpheusai.eth) no longer exists, so there is no Snapshot voting venue to observe. The Snapshot space this row confirmed no longer exists. Checked four ways on 2026-09-18, each with a positive control in the same request or the same UI so an absence cannot be an artefact of a changed API: the Hub's space(id:"morpheusai.eth") returns null, spaces(where:{id_in:...}) omits it while returning morpheusswap.eth from the same call, name searches for "morpheus ai", "MOR" and "mrc" surface no Morpheus AI space, and snapshot.box renders "Failed to load space" for it while rendering MorpheusSwap's 46 proposals normally. The MRC process and its public repository are a separate matter and are not claimed here; what is no longer verifiable is the Snapshot voting venue, which is the half of the claim this row was built on. Queried the Hub by id and by id_in, searched by name, and rendered the space in a browser, because a snapshot.org URL returns HTTP 200 for any id: it is a single-page app and the shell always loads. A control was included in every step after a first pass nearly produced a false alarm: aave.eth also returns null, not because the API is broken but because Aave's space is aavedao.eth.

◆Established ◑ Cross-checked Checked four ways on 2026-09-18, each with a positive control. The MRC process and its public repository are separate and not graded here. Whether MRC voting has moved elsewhere is a question for the team and is flagged for the monthly review.

The Base Builders proxy points at the source-verified BuildersV4 implementation, and the slot is watched for change. The EIP-1967 implementation slot on the Base Builders proxy 0x42BB446eAE6dca7723a9eBdb81EA88aFe77eF4B9 points at 0x18faef315b40a6d9cf49628f1133b1aa507513b0, source-verified on BaseScan as BuildersV4. Direct storage read of slot 0x360894...bbc, stored verbatim as an identity string and compared for exact inequality by check:primary-drift on every run.

◆Established ● on-chain This is the artefact behind the no_central_control verdict: the 5-of-9 Safe can repoint this slot with no timelock, so which implementation it points at is the thing that has to stay watched. The watch was blind until 2026-08-17. The slot was hex-parsed to a JavaScript number, which cannot hold 160 bits, so the stored value round-tripped to 0x18faef315b40a700000000000000000000000000 and the drift guard was comparing addresses by percentage; a replacement landing within 10% of the current address would not have fired. Now stored as an opaque string and compared for exact inequality, so any repoint raises an ERROR.

MOR has a broad holder base, counted directly from every MOR Transfer event on Base. 14,752 addresses hold MOR (balance above zero), from our own Alchemy enumeration of every MOR Transfer log since deploy, balances reconstructed locally. getLogs Transfer enumeration + net-balance reconstruction, count(balance > 0); computationally independent of MorScan.

◆Established ● on-chain Not score-bearing.

Provider registration has no allowlist, blocklist or approval step; the only owner lever on the registry facet sets the minimum stake for everyone at once. Nothing in the registration path can reject a specific address. ProviderRegistry.providerRegister on facet 0xE30279b79392AEfF7fDf1883C23d52eBA9D88A75 (source-verified on BaseScan) is unmodified by onlyOwner and reads no allowlist, blocklist or attestation; the only owner function on that facet is providerSetMinStake, which moves the capital floor for everyone at once. We registered our own node from a fresh address with no approval step (tx 0x1b8795ee52fd1fc6646edd58dc98c76c182999b73a641b90772e38cfb6390f8b, 1 September 2026) and it reads back active. Re-verified 2026-09-18 from the deployed LumerinDiamond (0x6aBE1d28, Base): the full 73-selector facet map was enumerated through the diamond loupe and providerRegister 0x365700cb is present in it, which is the same read that served as the positive control for the content-policy check. Registration remains permissionless: no owner modifier and no allowlist gate the selector, and we registered our own provider from a fresh address with no approval step. Resolved the providerRegister selector 0x365700cb through the Diamond loupe to its facet, read the verified source for allowlist and blocklist checks, then registered a provider from an unapproved address and confirmed the ProviderRegistered log.

◆Established ● on-chain Scoped to registration. Whether a live session can be censored once a provider is inside is the separate no_protocol_content_policy check, and one of the five facets (0x3a3952F0) has no verified source, so that one stays open.

The deployed inference contract has no pause, blacklist or content-policy function among its 73 selectors. Supported at the contract layer. The deployed LumerinDiamond on Base (0x6aBE1d28) exposes 73 function selectors across five facets, read from its own loupe on 2026-09-17. None of them is pause(), unpause(), paused(), setPaused(bool), a blacklist or ban setter or getter, or a content-policy setter. The two removal functions that do exist are marketplace hygiene rather than moderation: providerDeregister(address) reverts ProviderNotFound() for a caller unrelated to the provider, so it is a self-service exit and not an admin lever, and modelDeregister(bytes32) reverts a custom error naming the caller. There is no function through which the protocol could filter inference by content. Enumerated the COMPLETE selector map via facets() rather than testing candidate selectors one by one, because a Diamond reverts identically on a gated function and on one that does not exist, so a revert proves nothing on its own. Positive control: providerRegister 0x365700cb, the selector our own permissionless_provider verdict resolved through this Diamond, is present in the returned map, so the read is discriminating. Candidate moderation selectors were then checked for MEMBERSHIP of that map, and the two deregistration functions were called from an unrelated address to see which error they return.

◆Established ● on-chain Performed 2026-09-17, having been parked since 2026-08-09 with the method written down and never run. SCOPE: this grades the contract layer, which is the layer the claim is about. A Diamond can add a facet, so the finding is a statement about today's selector map and the watcher now re-probes the Diamond daily, where an owner change or an upgrade routes to a human. Provider-level self-censorship is off-chain and separate, and the proxy-router that providers run can filter whatever its operator chooses; neither is claimed here, and neither would show in a selector map.

Users and providers act from keys they hold; no custody contract sits between a user and their MOR, and session collateral is escrowed and returned at close. Supported, from our own operation of the protocol rather than from its documentation. We registered a provider on the deployed LumerinDiamond from a fresh address whose key we hold, with no approval step and no account creation, and we run sessions from it: the finding recorded in the permissionless_provider row on 2026-09-07. No custody contract sits between a user and their keys. The one thing the protocol does hold is session collateral, which the contract escrows for the life of a session and returns at close, and that is collateral rather than custody. Operated the protocol from a key we hold, which is the only way to test a custody claim: a contract read can show the absence of a custody function, but not that a user can actually transact without surrendering a key to anyone.

◆Established ● on-chain Performed 2026-09-17. It had sat unverified since 2026-08-09 carrying the note "Likely match", which is the shape of an answer nobody had gone and got. SCOPE: this grades key custody, which is what the claim says. Session collateral held by the contract during a session is escrow and is graded where it belongs, in the payout and session rows.

Morpheus's core code is public on GitHub, but a large share of its repositories carry no licence file. Re-read 2026-09-18 through the GitHub API: 50 non-fork public repositories in github.com/MorpheusAIs (63 public in total, 13 of them forks), 21 carrying no licence file and 34 not pushed in 180 days, newest push dashboard-v2 on 2026-09-15. The count is stated on a non-fork basis, which is lower than the public total this row quoted before: counting forks credits a project for code it did not write. GitHub API confirms the public repo set and MIT licensing. Repo count + license read from the org.

◆Established ◑ Cross-checked Counted on a non-fork basis, because counting forks credits a project for code it did not write.

The Builders proxy on Base points at a BuildersV4 implementation that is source-verified on BaseScan. The Builders proxy (UUPS, Base 0x42BB446e) implementation slot reads 0x18faef315b40a6d9cf49628f1133b1aa507513b0, source-verified on BaseScan as BuildersV4 (Solidity 0.8.20, exact match). The LlamaAI verification gap is resolved. eth_getStorageAt(0x42BB446e, EIP-1967 impl slot) -> 0x...18faef...; BaseScan shows Source Verified / Exact Match, ContractName BuildersV4.

◆Established ● on-chain Closes the LlamaAI Builders-contract forensic: the live implementation is verified as BuildersV4. Residual risk (graded under governance): UUPS upgrade authority is a 5-of-9 Safe with no timelock, so a future upgrade could swap to an unverified implementation - monitored by a check:primary-drift ERROR on the impl slot. Moved to the control-surface watch on 2026-09-18: the EIP-1967 implementation slot this row reads is probed daily on the Builders contract, so an upgrade that would invalidate "source verified at this implementation" routes to a human the next morning instead of waiting for a re-read.

Builder staking is live and carries material deposits on Base and Arbitrum. 2,669,995 MOR deposited across builder subnets ON BASE, read from totalDeposited on the 0x42BB Builders contract in our own harvest, re-read 2026-09-18. The scope matters and had not been stated: this is the Base leg only, and the Arbitrum Builders contract holds a further 397,598 MOR as of 2026-09-10, so the network total is nearer 3,067,593. Base deposits jumped about 34% in the twelve days to 2026-09-18, from 1,996,002 on 6 September, which is real growth rather than a re-basing: the series carries every intermediate reading. Sum of per-subnet totalStaked across our harvested list; MorScan totalDeposited as the cross-check anchor (its allPoolsTotal 9.77M was rejected as undocumented).

◆Established ● on-chain Read from the Builders contracts in our own harvest; the Base leg is banded. Re-anchored on 2026-09-18.

Inference usage is attested on-chain: every closed session writes a token receipt to the inference contract, and our own index sums them. 37.1B attested input+output from our own index of the LumerinDiamond SessionClosed receipts on Base, the same contract Morpheus indexes. The anchor read 10.6B when this verdict was first authored in June 2026 and has climbed every month since. sum of provider-signed SessionClosed receipt inputTokens+outputTokens; the legacy gap reproduces Morpheus’s documented clamped fallback.

◆Established ● on-chain Wired to our own index so the figure cannot fossilise. Never cite the 33.8B estTokens figure, an estimation artefact of ours (see the estTokens_legacy_overcount flag).

Providers post live price bids on-chain across most active models. The contract carries 729 active bids across 394 models, out of 407 active models and 461 registered, enumerated from the LumerinDiamond on Base. Per-model enumeration summed across the active model set, counting returned ids rather than trusting a total field. There is no global bid enumerator on the Diamond.

◆Established ● on-chain A bid count is a supply-side signal and says nothing about whether anyone buys at those prices.

Sessions settle on-chain: our own index counts every SessionClosed event on the inference contract. 358,715 sessions closed, from our own index of LumerinDiamond SessionClosed events on Base. Enumerate SessionClosed receipts on the LumerinDiamond and count them; previously cross-checked against a full Alchemy getLogs enumeration of SessionOpened and SessionClosed.

◆Established ● on-chain A raw closed-session count, including early-terminated sessions.

Value accrues through yield on capital deposits routed to the protocol; there is no fee burn. DeFiLlama summary/fees/morpheus-ai (dailyFees), re-read 2026-09-19: $557,653 trailing-year and $9,910,716 all-time. The trailing year has fallen every reading: $980K when the claim was written, $737,188 in August, $620,726 on 18 September, $557,653 a day later as another high-revenue day aged out of the window. The 30-day figure is $43,819, annualising to about $533,131. adapter tracks stETH rebasing + Aave V3 interest on capital-provider deposits.

◆Established ◑ Cross-checked Revenue falls with TVL. Banded on the trailing year, which can fall, rather than the all-time total, which only rises.

No vesting, cliff or team-allocation contract holds MOR back, so total supply is circulating supply; burned MOR at the dead address is still counted in totalSupply. Live totalSupply() summed across the three deployments on 2026-09-06: Ethereum 68,924.52 + Base 5,880,363.48 + Arbitrum 3,348,173.39 = 9,297,461.39 MOR. No vesting, cliff or team-allocation contract exists to hold a tranche back, which is what the claim is about. One qualification the claim does not make: 287,795.42 MOR sits at 0x...dEaD on Arbitrum and totalSupply() still counts it, so total equals circulating only if burned tokens count as circulating. Summed the three legs in one run, because a bridged token's supply is the sum and no single leg is the total. Burn balance read separately on the chain that holds it. Ingestion only, verdict human-set.

◆Established ● on-chain No locked tranches; total equals circulating. DEEPENED 2026-09-06: the verdict previously asserted the cross-chain read without stating a figure or re-reading it, so nothing would have noticed if a leg had changed. Now measured and wired. The burn qualification is the honest limit on the claim, and it is also where CoinGecko goes wrong: its total and its circulating each net the same 287,795.42 MOR, so its two figures differ by a burn that has already been subtracted once.

Protocol revenue from capital-deposit yield is falling: the trailing year keeps shrinking as higher-revenue days age out of the window. DeFiLlama summary/fees/morpheus-ai (dailyFees), re-read 2026-09-19: $557,653 trailing-year and $9,910,716 all-time. The trailing year has fallen every reading: $980K when the claim was written, $737,188 in August, $620,726 on 18 September, $557,653 a day later as another high-revenue day aged out of the window. The 30-day figure is $43,819, annualising to about $533,131. Trailing-12m fee total from the open adapter.

◆Established ◑ Cross-checked Same source as protocol_yield_revenue. Sustainability is the separate question in demand_economic_not_theatre.

The compute emission pool fell from its mid-July 2026 peak while a contract flaw let stake be reposted the same day, and has refilled since the 31 July fix: it now sits above that peak. 2,862,944.8 MOR on 2026-09-23 (LumerinDiamond.getComputeBalance on Base), above the ~2,805,000 mid-July peak, and rising ~2,644 MOR/day: accrual ~2,889/day against claims ~245/day over a 7-day window read from chain. The July drawdown coincided with the stake-recycling flaw fixed by the DiamondCut at block 49367526 (31 July 2026). getComputeBalance read at archive blocks 30/60/90 days back, differenced against live accrual.

◆Established ● on-chain Years of runway remain at the current rate. Corrected 2026-09-24: the row still said the pool was falling ~899 MOR/day, a pre-fix reading; the level stayed within the freshness band while the direction reversed, so no crossing fired.

Capital deposits in the Distribution contracts are the principal the protocol's yield is earned on. $20,698,328 protocol TVL across Morpheus chains, re-read 2026-10-07 ($24,191,115 on 2026-09-24). DeFiLlama protocol TVL summed across the chains its morpheus-ai adapter covers, refreshed into meta.primary_data.

◆Established ◑ Cross-checked Carried beside protocol_yield_revenue: a sustained fall here shows up as falling revenue later. Same aggregator as the revenue leg, so the two move on a consistent basis. Re-anchored 2026-09-18.

Provider payments settle from a funded multisig account that is topped up by manual action and spent down as sessions close. 93,522 MOR held by the compute funding Safe 0x5160C0311A95E0A1072FA85Df23712A7BA1cD4b1 (Gnosis 5-of-9), re-read 2026-09-29, down from 105,826 on 2026-09-18 and 122,996 in August. The balance falling is the account being spent down to settle sessions, which is the mechanism working rather than failing, and it is also the runway shortening. Direct ERC-20 balance read, refreshed into meta.primary_data and scaled to whole MOR.

◆Established ● on-chain The near-term constraint the pool figures hide. The compute pool is the long-dated subsidy; this Safe is the account closeSession actually pulls from, it is topped up by manual multisig action, and if it runs dry closeSession reverts and user stake stays locked. That makes a falling balance a liveness signal rather than a tokenomics one, which is why it is wired separately from compute_subsidy_sustainability. Stored in wei until this pass, so the drift band was reporting in units nobody reads; converted to whole MOR here, snapshot and series together.

MOR trades only in DEX pools, and those pools are thin. DEX pool depth is on-chain-measurable; confirms thin liquidity. DEX pool depth read on-chain.

◆Established ● on-chain Thin DEX liquidity, hence the low dimension score (4/15). Re-checked 2026-09-18 against the venue read added to this project the same day: CoinGecko lists six venues for MOR and every one is a DEX pool (Uniswap V3 on Arbitrum, Uniswap V4 and V3 on Base, Aerodrome), which is the on-chain pool depth this row describes. The count is wired on the sibling cex_liquidity row, so a new venue routes there.

No centralised exchange lists MOR; every venue CoinGecko tracks is a DEX pool. Six venues listed on CoinGecko as at 2026-09-18, and every one of them a DEX pool (Uniswap V3 on Arbitrum, Uniswap V4 and V3 on Base, Aerodrome). No centralised exchange carries MOR, which is what this verdict says. CoinGecko tickers/volume confirm limited CEX listings and thin volume. Read tickers + volume.

◆Established ◑ Cross-checked Off-chain, api grade.

Inference demand is served on-chain, but monthly tokens processed fell in August 2026 after six months of growth. Our LumerinDiamond index, recomputed 2026-09-05 with August complete: monthly tokens processed rose every full month February to July 2026 (1.72B, 1.89B, 2.33B, 3.00B, 7.12B, 16.12B), then FELL to 9.96B in August, -38.2% month on month. The 2026-08-17 assessment projected ~10B for August from 16 partial days and that projection landed (9.96B actual). Sessions fell further than tokens over the same step, 83,699 to 36,286 (-56.6%), so tokens per session rose ~42% (192,574 to 274,374): fewer but materially larger sessions rather than a uniform contraction. The growth half of the claim no longer holds on the last complete month; the 'real demand is served' half still does. Monthly sums of provider-signed receipt input+output tokens across the full series, plus session counts for the same buckets. Cross-check: tech.mor.org's monthly 'tokens processed' chart returns 16,118,221,761 for July and 9,955,935,085 for August, byte-identical to our index, and its lifetime session total (338,643) sits within 44 of ours (338,599), read minutes apart. Both derive from the same SessionClosed events, so this confirms our extraction rather than providing an independent source. NOTE the segmentation differs: tech.mor.org splits Native vs API Gateway (Jul 1.02B/15.09B, Aug 4.25B/5.70B) while our nativeTokens field counts only MOR-paid tokens and is near zero. Do not read our split as theirs.

◆Established ● on-chain Reproduced from our own index of the same contract Morpheus indexes. The cumulative anchor is wired to catch the next move.

A 4-of-6 Safe controls VVV emissions and staking-logic upgrades, and no distributed governance is deployed. A 4-of-6 Safe controls emissions and staking-logic upgrades; no distributed governance exists. Same on-chain control trace as the no_token_governance row: VVV.owner() is StakingV2, and StakingV2.owner() is the Safe.

◆Established ● on-chain Venice's launch post says the company is owned exclusively by its team, and the token carries no governance rights (see no_token_governance).

VVV is held across a wide base, but the ten largest non-infrastructure holders still hold close to a third of supply. Measured on Base 2026-09-18 across 150,671 holders: the top ten hold 63.37% of supply raw, and 30.72% once infrastructure addresses are excluded, with 33.17% of supply sitting in infrastructure (staking, treasury, liquidity, burn) and 26.97% of the raw top ten identifiable as insider. Four of the top ten are unlabelled. So the airdrop-led launch did produce a wide base, and a third of the float still sits in ten hands once the protocol's own contracts are set aside. Top-10 holders with infrastructure addresses classified and excluded from the numerator while their balances stay in the denominator, which is the arithmetic that matters: excluding a burn or staking balance from both sides is how OCEAN's 81%-at-0xdEaD once made an infra-excluded share read five times lower than the live-holder share.

◆Established ● on-chain Four of the top ten addresses are unlabelled. Measured with the same holder index used for nineteen other tokens.

The VVV token contract has no pause, freeze or blacklist function, and it is not upgradeable. Verified VVV source exposes no pause, freeze, or blacklist function; owner-gated functions are mint + transferOwnership only, and the contract is not upgradeable (no EIP-1967 slots). Enumerated the verified ABI for pause/blacklist/freeze; read EIP-1967 implementation/admin slots (both zero = immutable).

◆Established ● on-chain A token-layer property only. Inference-layer content policy is covered in uncensored_inference.

Most of Venice's text models are open-weight models it hosts itself; the rest are closed frontier models proxied to outside providers. Read from Venice's own public model API on 2026-09-17: 118 text models, of which 69 are Venice-hosted open-weight (model_spec.privacy "private") and 49 are closed frontier models proxied to an external provider ("anonymized"), including every Gemini entry. Three of the 69 are explicitly uncensored builds (venice-uncensored-1-2, venice-uncensored-role-play, and an abliterated GLM-4.7 variant). Recall resistance holds for the 69 Venice-hosted models and does not for the other 49, which a third-party provider can withdraw at any time. Counted model_spec.privacy across the full catalogue. The field discriminates per model rather than defaulting: "private" is Venice-hosted, "anonymized" is a proxied closed model.

◆Established ◑ Cross-checked This covers the model list; whether switching Safe Mode off removes content filtering is a setting inside a closed-source proxy that no outside read can confirm. Venice controls that proxy and already gates by jurisdiction (the free tier is blocked in the UK under FSMA), so censorship resistance here is a business decision that 69 downloadable model weights make expensive to reverse.

The VVV token and StakingV2 contracts carry verified source on BaseScan. VVV token and StakingV2 both show verified source on BaseScan (Solidity 0.8.26). Queried getsourcecode for VVV and StakingV2: both verified.

◆Established ● on-chain On the daily control-surface watch since 2026-09-18, which reads source verification and the contract name from Blockscout. Reading "Venice" back confirms the probe is on the right contract, and a proxy upgrade to an unverified implementation would show as a diff.

Venice's core inference engine and proxy are closed source; only peripheral repositories are public. Core inference engine, proxy, and platform code are closed-source; only peripheral repos are public. Privacy claims cannot be independently verified because the proxy code is not published. Reviewed the public repo set against the described architecture; the load-bearing inference/proxy code is not among them.

◆Established ◑ Cross-checked Open-source rubric stage O1: peripheral code, verified contracts and one audit, with the core closed and the Trust Security report not publicly downloadable.

VVV has no supply cap: the contract has no cap function and issuance is an owner-gated mint. No cap() on the contract; owner-mint issuance confirmed; CoinGecko max_supply null. Live effective supply 80,955,440 (on-chain totalSupply 114,809,304 minus 33,853,864 burn), re-read 2026-09-08; was 80.75M at the 2026-08-09 assessment. Read totalSupply and balanceOf(0x0) live; confirmed no cap function and owner-gated mint; reconciled effective supply against CoinGecko.

◆Established ● on-chain Supply rubric: uncapped and emission-funded, with a steep disinflationary trajectory and a burn offset. Venice's FAQ answers the supply-cap question with its emission schedule rather than a cap. Burns are transfers to 0x0 and do not reduce totalSupply, which makes a totalSupply delta a clean read of emissions (see emission_rate).

VVV trades on major centralised exchanges and Base DEX pools across dozens of venues. 75 tickers on 49 venues, re-read 2026-09-29 (66 earlier); Coinbase the top venue, with Bybit, Gate, Kraken and OKX alongside Aerodrome DEX depth on Base. Read the ticker set and 24h volume live; confirmed CEX and DEX venues.

◆Established ◑ Cross-checked Multi-venue liquidity. Structural language only in editorial (turnover %, % from ATH).

Akash's validator set is open but concentrated: five validators hold enough stake to halt the chain. 84 bonded validators against a 100-validator cap, securing 91,466,381 AKT (30.8% of total supply bonded). Stake is concentrated: the largest validator holds 11.5%, the top 5 hold 37.3% and the top 10 hold 60.1%, giving a Nakamoto coefficient of 5 (five validators can halt the chain). Enumerated the full bonded validator set via Cosmos REST and computed the stake-share distribution and the >33% Nakamoto coefficient directly from validator tokens. Ingestion only, verdict human-set.

◆Established ● on-chain Recorded as a measured anchor. Validator decentralisation is separate from compute-provider decentralisation, which this row does not cover.

Token-holder governance controls a funded community pool, fed by a 70% tax on new issuance. 5,702,854 AKT in the x/distribution community pool. Direct Cosmos REST read of the community pool balance for uakt, refreshed into meta.primary_data.

◆Established ● on-chain The scale check on governance_authority: binding votes matter to the extent there is something to vote over, and single proposals have asked for a large fraction of this balance. Proposal 322 raised the tax on issuance from 50% to 70%, which grows the pool while shrinking staker take. Re-anchored 2026-09-18 after the band flagged the move from 4,365,024 to 5,702,854.

AKT's liquid holder base is broad: over a million accounts hold a balance, and the ten largest accounts outside the chain's own modules hold under a third of supply. 1,136,071 accounts hold a non-zero AKT balance, every one of them enumerated rather than sampled. Ten accounts hold 63.10% of the 298,014,361 AKT bank supply, but three of those are the chain's own module accounts (bonded_tokens_pool 30.78%, not_bonded_tokens_pool 4.74%, distribution 3.44%), leaving an infra-excluded top ten of 29.48%. The largest non-module account holds 7.78% and the next six sit between 3.14% and 2.09%. Seven of the ten carry no label anywhere Akash publishes. Whole-owner-set enumeration rather than a rich list, because denom_owners is unsorted and a top ten is otherwise unobtainable. Module accounts are excluded as infra from the chain's own auth module; Cosmos publishes no explorer labels, so every other account stays unattributed. Ingestion only, verdict human-set.

◆Established ● on-chain This is the liquid holder base. A delegation leaves the delegator's bank account and sits in bonded_tokens_pool, which is why that module is the largest single line and why a staker's economic position can be far larger than the balance counted here. The denominator is the chain's own bank supply (298.0M) rather than the 391.2M our tokenomics block carries from CoinGecko; the two measure different things. Score held at 8/15.

A handful of Akash validators could halt the chain: five together hold more than a third of stake. Re-read 2026-09-18 from the chain's own endpoint: 82 active validators, halt coefficient 5 (the fewest exceeding a third of stake), top-10 share 60.88%. Consensus set read from each chain's own endpoint by scripts/refresh-validator-sets.ts: active validators, stake shares and the NAKAMOTO HALT COEFFICIENT, which is the fewest validators whose combined stake exceeds one third. That is the BFT liveness threshold on these chains and it is NOT the 51% control figure the two get confused for, which is why the output carries nakamoto_basis. Censorship needs the halt number rather than the set size: a chain with a hundred validators and a coefficient of five is five colluding parties away from stopping.

◆Established ● on-chain The Nodekeeper and jurisdiction arguments in the review stand. The chain adds that the transaction layer beneath them has a halt threshold in single figures, with under a third of supply bonded to defend it.

No AKT has been permanently destroyed by the burn-and-mint vault since it activated; AKT moves into the vault and back out to providers. 0 AKT permanently burned (x/bme vault_state.total_burned, denom uakt). Direct Cosmos REST read of vault_state.total_burned for uakt, refreshed by scripts/refresh-primary.ts into meta.primary_data.

◆Established ● on-chain Carried as its own row because every reading of Akash as burning AKT turns on it, and it has read zero since BME activated in March 2026. AEP-76 specifies net destruction only when AKT appreciates between vault top-up and provider settlement, so a permanent burn depends on price moves rather than usage. A first non-zero reading crosses every band.

Providers redeem most of the ACT credit the vault issues, so the burn-and-mint loop closes rather than piling up unspent credit. 799,299 ACT burned (redeemed) against 968,221 ACT minted since activation, so 82.6% of issued credit has been claimed. Direct Cosmos REST read of vault_state.total_burned and total_minted for uact; the share is the ratio of the two refreshed values.

◆Established ● on-chain The mint counter alone would not distinguish a working loop from one issuing credit nobody redeems. Redeemed ACT tracking most of minted ACT is the evidence that providers are actually settling through BME. The residual is credit issued but not yet claimed, which is expected for in-flight leases.

AKT has no supply cap: the mint module keeps issuing with no ceiling parameter, and the 388.5M genesis-schedule figure is not enforced on-chain. On-chain bank supply reads 296,747,666.10 AKT and the x/mint module continues to issue at 4% annually with no cap parameter; nothing in the module enforces 388,539,008. The on-chain figure matches CoinGecko's 296,746,575 total supply to within 0.0004%. Direct Cosmos REST read of the base-denom bank supply (uakt, 6 decimals) via the coldstart cosmos source, cross-checked against the aggregator figure. Verdict human-set.

◆Established ● on-chain Aggregators carry 388,539,008 as max supply; it is the genesis schedule target, and at 4% inflation the chain passes it inside a decade.

Tenants pay lease fees on-chain, and the trailing-year total runs in the low millions of dollars. $1,983,371 trailing-year / $6,211,646 all-time on-chain lease fees (DeFiLlama akash-network adapter, read 2026-09-20). DeFiLlama sums on-chain lease fees paid by users (revenue = AKT burned). Token Terminal lists Akash but exposes no revenue metric.

◆Established ◑ Cross-checked Banded on the trailing year, which can fall, rather than the all-time total, which only rises. Token Terminal lists Akash but exposes no revenue metric.

AKT trades on major centralised exchanges including Coinbase, Kraken and KuCoin at moderate turnover; Binance does not list it. Coinbase, Kraken and KuCoin all quote AKT across 30 tickers on 19 venues, re-read 2026-10-07 (34 on 2026-09-29), and no Binance market appears anywhere in the set (Binance exchangeInfo returns AKTUSDT as an invalid symbol, re-checked 2026-10-07). Upbit leads at 28.7% of 24h volume, then HTX at 16.8% and Coinbase at 13.1%, with Bithumb and Paribu adding Korean and Turkish access. Paginated the CoinGecko ticker set and summed 24h volume per venue; turnover from the trailing 7-day median of volume over market cap.

◆Established ◑ Cross-checked The absence of a Binance listing is the part most likely to change. The Community Pool market-making loan noted in the evidence is a separate matter and is not tested here.

Paid lease demand on Akash is concentrated: one payer accounts for most active lease spend. Grouping all 1,454 active leases by payer gives 529 distinct payers and total paid demand of about 11,675 ACT a day, of which the single largest holds 72.5%. That address (akash1scs6vfgwamnnarmsz26yy63244pakvul5rszxs) is topped up from one custody account (akash1226a4mhmdseh3mz34rslc0ud4pc9k4dwnn2d83), which is itself fed by three upstream accounts, a pattern consistent with a funded corporate tenant. Setting that payer aside leaves about 3,209 ACT a day across the remaining 528. Nothing on-chain identifies the custody account as Overclock Labs or AkashML. The second-largest payer (19.9%) receives no MsgSend from that custody account or any other, so it does not share the funding path. Full enumeration of active leases, summing price-per-block by lease.id.owner and converting at a measured block time (5.895s). Funding paths for the top three payers traced through cosmos/tx event queries on transfer.recipient. Lease prices are uact, so daily figures are dollars without touching AKT's price. Attribution of any address to a named company is NOT derived here.

◆Established ● on-chain Measured from every active lease grouped by payer. The attribution of the largest payer to AkashML comes from outside analysis (an X thread of 25 Jul 2026) and is not derived from our reads. Flagged for the Returns Sustainability review.

Eight validators together hold more than a third of NEAR stake, enough to halt the chain, out of an active set of over four hundred. 415 active validators, and the Nakamoto halt coefficient is 8: that many colluding validators exceed one third of stake and can stop the chain. The top ten hold 39.62% of stake, and 47.33% of supply is bonded. Consensus set read from each chain's own endpoint by scripts/refresh-validator-sets.ts: active validators, stake shares and the NAKAMOTO HALT COEFFICIENT, which is the fewest validators whose combined stake exceeds one third. That is the BFT liveness threshold on these chains and it is NOT the 51% control figure the two get confused for, which is why the output carries nakamoto_basis. Censorship needs the halt number rather than the set size: a chain with a hundred validators and a coefficient of five is five colluding parties away from stopping.

◆Established ● on-chain OYM holds NEAR. The widest validator set and the highest halt coefficient of the five chains measured here, with close to half of supply bonded.

NEAR's reference client, nearcore, is public under GPL-3.0 and actively developed, inside an org of 289 public repositories. Confirmed. The github.com/near org carries 289 public repositories; the reference client nearcore is GPL-3.0, 2,610 stars, not archived, and was pushed the same day we ran the check with 484 open issues. Direct GitHub org/repo metadata reads via the coldstart github source. Ingestion only, verdict human-set.

◆Established ◑ Cross-checked Covers licence and development activity, not audit coverage. The band re-measures stars, which prompts a re-look rather than proving the repo is still open: archiving or relicensing would not move it.

Aggregators disagree on NEAR's circulating supply by about 4% of total supply, and no holder of that gap has been identified. The two aggregators disagree by ~54.0M NEAR. CoinGecko reports circulating 1,303,818,416 against its OWN total of 1,303,818,394, so its circulating figure exceeds its total and it plainly models no lock-up. NearBlocks reports circulating 1,249,836,992 of 1,303,845,705, i.e. 95.86% circulating and 54,008,711 NEAR non-circulating. Chain total supply confirmed on two independent RPC endpoints. Read total supply from the chain, then compared each aggregator's circulating basis against it and against itself.

◆Established ● on-chain OYM holds NEAR. Neither aggregator basis is asserted as correct; CoinGecko's circulating figure exceeds its own total, so it models no lock-up. The gap connects to the foundation_treasury row. Supply Dynamics is flagged for review.

NEAR's base-chain gas revenue is small and has fallen steadily since June 2026. $1,271,551 trailing-year and $18,013,881 all-time protocol revenue, re-read 2026-09-18, down from $1.49M trailing-year in August and ~$2.0M in June. The 30-day window is $39,266, annualising near $477,736, so base-chain gas revenue is falling steadily rather than having stepped down once. DeFiLlama summary/fees/near?dataType=dailyRevenue (dimension adapter); chain TVL via /v2/chains. Ingestion only, verdict human-set.

◆Established ◑ Cross-checked Measured from DeFiLlama's open-source adapter for base-chain gas; Token Terminal's revenue series agreed with it to about 1% on a June read. Banded on the trailing year, which can fall, rather than the all-time total, which only rises. The Intents legs are measured separately.

NEAR Intents settlement fees are growing, with the 30-day run-rate ahead of the trailing year. $50.81M trailing-year and $52.73M all-time gross Intents fees, re-read 2026-10-07; 30 days $7.38M, which annualises to about $89.8M. Intents is the one leg of NEAR's revenue that is GROWING, and the 30-day run-rate is running ahead of the trailing year rather than behind it. DeFiLlama summary/fees/near-intents?dataType=dailyFees (Intents Verifier contract across chains); DISTINCT from base-chain `near` gas fees. Ingested into meta.primary_data by refresh:primary; verdict human-set.

◆Established ◑ Cross-checked Measured from DeFiLlama's open-source near-intents adapter, which reads the Intents Verifier contract across chains and is distinct from base-chain gas. Banded on the trailing year, which can fall. Re-anchored 2026-09-18.

DeFi capital on NEAR is tracked from DeFiLlama's chain series as a trailing 7-day median. $224,009,067 trailing-7-day median DeFi TVL on NEAR, re-read 2026-09-29 ($196,950,406 on 2026-09-24). DeFiLlama's daily series rose from about $140M on 15 September to about $204M on 23-24 September and has held above it since. DeFiLlama chain-TVL, refreshed by scripts/refresh-primary.ts into meta.primary_data. The stored figure is the trailing-7d median of the chain's own daily series. That was chosen against one- and two-day dropouts and does NOT survive the sustained flap seen from 15 August 2026, where three of five days read low: a median then reports whichever state dominates the window rather than the truth. Freshness banding is paused on this row until the adapter settles.

◆Established ◑ Cross-checked DeFi capital on NEAR, not Intents volume or AI activity, so it does not bear on the thesis this review rates. OYM holds NEAR, which is a reason to keep this measured. Banding was paused 2026-08-20 while the adapter's Near series alternated; lifted 2026-10-07 after the daily series held between about $168M and $249M from 18 September, read as a trailing-7d median.

NEAR is listed across a broad set of exchanges, including Binance, Coinbase, OKX, Kraken and KuCoin, though much of the reported volume sits on second-tier venues. All five confirm, and the volume floor is cleared several times over: 151 tickers on 89 venues, the widest coverage in our corpus, at $535.76M reported 24h volume and a 10.27% 7-day median turnover. The composition is worth stating alongside it. Binance leads at 16.2% and KuCoin follows at 10.0%, but BTCC, Toobit, CoinW and Websea occupy four of the next six places, so a material share of the reported flow sits on second-tier venues rather than on the tier-1 names the claim rests on. Paginated the CoinGecko ticker set and summed 24h volume per venue; turnover from the trailing 7-day median of volume over market cap.

◆Established ◑ Cross-checked OYM holds NEAR, so the second-tier share is stated alongside the listings. Liquidity is read as a turnover ratio rather than an absolute volume.

The top ten RENDER holders on Solana hold under half the SPL supply, with more than a fifth of supply in two multisig vaults. The top ten hold 48.53% of the Solana SPL supply, the lowest of the four measured, with 22.05% in two off-curve Squads-style vaults (16.89% and 5.16%) and 2.01% held by a PDA of the circuit-breaker programme that also gates minting. The rest is keypair wallets, one of which is Binance's Solana hot wallet at 4.54%. That last point is the basis caveat made concrete: on the EVM path exchange custody is excluded from the rated figure and here it cannot be identified automatically, so it is counted. Resolved each of the largest token accounts to its owner, then classified the owner on-chain rather than by label: System-owned and on-curve is a keypair wallet, System-owned and off-curve is a PDA with no private key (in practice a Squads-style vault), and a known pool program is neutral infrastructure. The rated figure is the RAW top-10 share. This is NOT the EVM ex-infra basis: exchange custody cannot be identified without labels and is therefore included, which is visible here rather than assumed.

◆Established ● on-chain Consistent with a token that migrated across chains and holds a large treasury. The two vaults are multisig-controlled rather than single-key, which the curve test establishes. Exchange custody cannot be identified on Solana without labels, so it is counted in the raw figure.

The RENDER mint carries a live freeze authority, and both programs behind its freeze and mint authorities are upgradeable by one Squads multisig with no timelock. The RENDER SPL mint has freezeAuthority set to a PDA of program distZXJ5FYrPhjBhB5P2BQ9B2AsPzJ4TcUSz6hKssP1 and mintAuthority set to a PDA of program circiqFCstNzaFBji1udQ6txgQBrn29pVSYHNJQo3wZ, so the token carries a live freeze lever. Both programs are upgradeable (BPFLoaderUpgradeable, upgrade authority present, not frozen) and both name the same upgrade authority, 7CVt936gVDXfKeXdRs5xcWVkrEaYGMTV3HA2K7j4Bqa7. That authority is a Squads multisig vault: System-Program-owned with no data, but off-curve, so no private key can exist for it; Solscan labels it 'Squad Vault "render"' and its history is vaultTransactionExecute calls. Replacing either program needs the multisig's threshold, with no timelock. This is the only one of the four assessed Solana natives (NOS/GRASS/AI16Z/RENDER) with a live freeze authority. Four-hop read: mint -> authority PDA -> owning program -> programData upgrade authority -> authority account shape. Shape now tested by ed25519 CURVE MEMBERSHIP, not owner+size: a Squads vault is System-owned and dataless exactly like a wallet, and only the off-curve test separates them. Automated by npm run watch:control-surface; multisig label attributed to Solscan.

◆Established ● on-chain A token-layer lever, separate from the network-layer censorship score. The multisig's signer set and threshold controls are not published. Both programs are on the control-surface watch list.