Auki Review

What it does

Auki Labs is a Hong-Kong-based spatial-computingSpatial ComputingComputing systems that understand and operate within the three-dimensional structure of a physical space. Combines computer vision, sensor fusion and 3D mapping so that software can reason about where things are, not just what they are.Regular computing tells you what's in an image. Spatial computing tells you that the cereal box is on shelf three, two metres from the till, with eight units left. The room becomes the user interface.Spatial Computing in the glossary → company founded in 2019 by Nils Pihl (CEO) and Santeri Aramo (COO). The team is small, around 32 people per LinkedIn, and the funding history is light: roughly $13.25 million across a $250,000 pre-seed and a $13 million seed led by Shima, Kenetic and Animoca in March 2022.

The product roster is two layers. Cactus is the B2B spatial-AI retail tool that gives shop staff a phone, smart-cart or robot-mounted view of the store’s inventory state, with an AI copilot that knows where each product sits. The posemeshPosemeshA decentralised machine-perception network built by Auki Labs. Domain servers host the 3D map of a venue, reconstruction servers turn scans into spatial data, and Hagall handles real-time networking between devices.A neighbourhood watch for physical spaces. Every venue owner runs a local server that knows the layout of their space; devices ask the nearest one for directions, and the protocol coordinates updates without a central operator owning the map.Posemesh in the glossary → is the underlying decentralised perception network: domain servers host a persistent spatial map per venue, reconstruction servers turn customer scans into 3D point clouds or gaussian splatsGaussian SplatA 3D rendering technique that represents a scene as millions of small fuzzy ellipsoids ("Gaussians") instead of polygon meshes. Each splat has position, colour and shape; rendering blends them together for a photorealistic view.Instead of building a 3D model out of LEGO bricks (polygons), you spray-paint it with millions of tiny coloured smudges that blur together when you step back. The result looks like a photograph rather than a video-game model.Gaussian Splat in the glossary →, and the Hagall layer handles real-time networking between devices.

The AUKI tokenTokenA digital unit of value or access rights tracked on a blockchain. Tokens can represent ownership in a project, a right to use a service, a share of future revenue, or simply a tradable asset with no underlying claim.Like a physical poker chip a casino issues. The chip itself has no value. What makes it worth something is what it lets you do at the casino, what the casino has promised, and how much other people will pay you for it.Token in the glossary → sits at the protocol layer. Retailers and developers burn AUKI to receive non-transferable off-chain service credits. Operators of reconstruction and domain nodes stake AUKI as reputation to be eligible for work and rewards. Cactus, the customer-facing retail product, is billed in USD, not AUKI. Hold that decoupling in mind through the rest of this review.

Cactus is the part that ships

The cleanest evidence-led story on the Auki page is Coop Stora Visby in Sweden. Coop Sweden ran a four-month Cactus deployment at the Stora Visby flagship, a 5,000-plus-square-metre store with 24,000-plus SKUs and dozens of employees. The published case study reports a minimum 15 minutes saved per user per day, one daily recurring shift-handover meeting eliminated, and staff productivity gains across pick-and-pack, replenishment and customer assistance. ESM Magazine carried a retail-trade write-up of the same deployment, which gets it past the self-reported-only bar.

Beyond Visby, the pilot roster is documented but smaller than third-party commentary suggests. Pepito in Bali is live. Reitan REMA 1000 in Denmark started a three-store pilot in 2025 with what Auki describes as potential to scale across 1,900 locations. FairPrice in Singapore began a pilot in January 2026. Auki’s Mars Inc partnership covers “thousands of new locations” under NDA with no public deployment count. A US Fortune 500 chain is named in community updates without disclosure of which.

Auki’s 5 September 2026 community update says it has signed a first contract with a gas-leak monitoring partner that already does this work across retail locations, with the first detection robot due “hopefully within a month”. The partner is unnamed and the deal unpriced, so this is Auki’s claim alone.

Here is the correction the DeAI triage system flagged. The “40,000 retail stores” figure that surfaced in commentary on 13 May 2026 is not in Auki’s own communications. It most likely conflates an unrelated Retail Dive headline about projected US store closures with Auki’s pilot footprint. Auki’s own figure is 6,500-plus retail locations representing the combined chain-footprint of pilots running paid Cactus, with Auki’s explicit caveat in the 1 December 2025 community update that “these aren’t all simultaneously active” and that “large retailers typically pilot across 3 to 30 locations”. The active deployment count is in the dozens. Treat any figure higher than that as marketing until primary-checked.

The product clearly works in the stores where it has run. The headline number being marketed in DeAI Twitter and triage docs is not Auki’s own.

The posemesh, currently centralised

The OpenZeppelin audit that covers AUKI’s five Solidity contracts also describes the posemesh as “currently a centralised system” with multi-party-computation wallets controlling privileged on-chain roles. Auki Labs has not pushed back on that framing in public. The 17 May 2025 community update used the phrase “a great deal of centralized control over the protocol in its current shape” and described progressive decentralisation as a roadmap. So the centralisation is admitted by both auditor and project.

The architecture is meant to decentralise. Reconstruction servers consume scan jobs from the Auki Domain Management Tool app and run a custom structure-from-motionStructure From MotionA computer-vision pipeline that reconstructs the 3D structure of a scene from a series of 2D photos taken from different angles. Used to build spatial maps for AR, robotics and decentralised perception networks.Walk around a sculpture and take twenty photos from different angles. Structure-from-motion is the algorithm that figures out where you were standing for each photo and uses that to reconstruct the 3D shape. The camera positions and the 3D model fall out of the same calculation.Structure From Motion in the glossary → pipeline. The Docker stack is open-source under the MIT licence (Python, Rust and C++). The hardware bar is a mid-tier NVIDIA GPU: 8 CPU cores, 12 GiB of RAM, 8+ GiB of VRAM, CUDACUDANvidia's parallel computing platform. The software layer that lets AI workloads run on Nvidia GPUs. Almost every serious AI model is trained and served through CUDA, which is why Nvidia has a structural moat in AI compute.Like electrical plugs. You can build any kind of appliance, but if every socket in the country is one specific shape, you have to match it. CUDA is that socket shape for AI, and Nvidia owns the factory that makes them.CUDA in the glossary → 12.8 and 10 Mbps of connectivity. The H2 2025 recap describes the reconstruction network as being in “closed beta with a handful of community node operators.” Domain servers are the lighter path and run on minimal CPU, but the same beta gating applies to participation.

The GitHub posture is open. The aukilabs organisation runs 55 public repositories, 37 of them original rather than forks, and 16 of those 37 carry no licence file (GitHub API read, 18 September 2026). The posemesh core repo is MIT-licensed with 47 stars and a last push of 11 August 2026, and at the May 2026 research date it had 474 commits from 12 contributors. The code-concentration caveat holds: 90% of the commits come from three contributors (shuning-auki, jnkvca, tatiesmars), which is consistent with a 32-person company and sits awkwardly beside the protocol’s decentralisation claims.

The audit verdict on the contract side itself is solid. OpenZeppelin reviewed AukiToken, BurnContract, RewardLiquidityPoolContract, StakingContract and UUPSProxy between 13 and 15 November 2023. Zero critical findings, zero high, two medium, two low and nine notes. Eleven of thirteen findings were resolved via PR #48. The two medium findings flagged are the kind that matter for an upgradeable token: the paused contract did not pause allowance changes (since fixed), and the DEFAULT_ADMIN_ROLE was assigned with the default unsafe pattern. Both are documented and addressed.

The risk that the audit does not fully resolve is the proxy posture itself. AUKI is a UUPS upgradeable proxyUUPS ProxyA pattern for upgradeable smart contracts where the upgrade logic sits in the implementation contract itself. Cheaper gas than the older Transparent proxy, but the implementation must be written carefully or the contract can be bricked.A power tool where the swappable head also contains the swap button. Pull the head off, snap on a new one with a new button, you can keep using the tool. But if the new head forgets to ship a swap button, the tool is now stuck on whatever head you just attached.UUPS Proxy in the glossary → at 0xf9569cfb8fd265e91aa478d86ae8c78b8af55df4 on Base. The implementation contract is 0x408Be34f38bE0ccD19F3acFE76C0b0Ff658EdFcF, last upgraded 14 July 2024. Pause, upgrade and an uncapped mint all sit behind a 4-of-6 Safe, so its key-holders can change the implementation. For a token that is supposed to clear retail commerce, that is a meaningful chokepoint.

Tokenomics

Auki’s whitepaper sets an initial mint of ten billion AUKI and an equilibrium target of five billion. Neither bound is enforced in code: the upgradeable token carries an uncapped mint behind a 4-of-6 Safe. Total supply on-chain is about 9.99 billion. Circulating supply per CoinGecko is approximately 4.43 billion, which is about 44% (CoinMarketCap’s lower figure appears stale).

Cumulative on-chain fees since launch are $254,298 per DeFiLlama, with $79,488 over the trailing year (October 2026 read, down from $102,983 in September, $146,781 in August and $166,400 in June). Annualised on-chain incentive emissions ran at roughly $371,000 when last measured in June 2026, so set against that figure net on-chain earnings are roughly negative $292,000.

The allocation table sums to exactly 100% per the whitepaper:

Category	%	Vesting
Ecosystem Rewards	30.00	36-month daily linear
Posemesh Foundation	18.74	84-month daily linear (longest in the schedule)
Team	15.58	42-month, 6-month cliff (opened 28 Feb 2025)
Token Infrastructure	7.00	Immediate at TGE (market-making and liquidity)
Pre-sale 1 + Pre-sale 2	6.50	24-month daily linear (Pre-sale 1) / 18-month (Pre-sale 2)
Accelerator (Outlier Ventures)	6.00	42-month, 6-month cliff
Advisors	4.93	36-month daily linear
Seed	3.51	48-month daily linear
Pre-seed	1.09	48-month daily linear
Early Bird	4.50	18-month daily linear
Community / KOL Pre-sale	2.15	Shorter daily linear schedules

On paper the vesting is unusually disciplined for a 2024 launch. The whitepaper puts each allocation on a daily linear schedule, with no day-one major TGE release beyond the 7% Token Infrastructure tranche that seats market-making, and no cliff longer than six months. Combined insider weight (team plus accelerator plus seed plus pre-sale tranches) sits in the 25-to-35% range depending on definition, which is below the typical 40-to-50% range for comparable 2024 launches.

Code enforces little of that schedule. The top 10 addresses hold 83.8% of supply: 79.74% in eight plain wallets that trace to the token deployer, and only 3.01% in the MerkleVester contract (24 September 2026). So the release schedule rests on Auki’s word.

The economic mechanism is a burn-for-credit-mint loop. A retailer or developer wants Cactus service credits, denominated off-chain in USD. They burn the dollar-equivalent of AUKI into the BurnContract. The protocol records the credit and triggers a deflationary mint that is mathematically smaller than the burn. Net effect: supply trends toward the five-billion equilibrium. Holders earn nothing from the mechanism; the value claim is that as throughput grows, the supply curve compounds against demand.

The flywheel is currently trivial. With trailing-year on-chain fees small and falling, the burn-mint mechanism is not yet meaningfully reducing supply. The deflationary structural claim is sound. The flow that would activate it is absent.

The price action reflects the cap-table arithmetic. AUKI printed an all-time high of $0.1812 on TGE day, 28 August 2024 (CoinGecko records its own high lower, at $0.0606 on 26 January 2025, so treat the TGE-day print as a thin-liquidity spike). The current price sits roughly 97% below the TGE-day high. The all-time low of $0.004161 was set on 5 February 2026. The MCap/FDV ratio is about 0.44. Twenty-four-hour volume is thin against the cap.

The token is bridged to peaq via LayerZero (Stargate and Superbridge), and the AUKI/PEAQ pair trades on MachineX, the peaq ecosystem DEX. That cross-chain footprint is the strongest piece of the access story: it gives AUKI a foothold in the machine-economy DEX network that peaq has been building.

How to participate

Operating a reconstruction node is open-source (MIT, Docker stack) but the network is in closed beta. The hardware bar is a mid-tier NVIDIA GPU; AUKI reputation stake is required to earn rewards, but the specific stake amount is not publicly documented at research date. Node deployment runs through the NuNet Appliance per the December 2024 Posemesh x NuNet partnership announcement, with NuNet’s Device Management Service handling capability matching and on-chain settlement.

Operating a domain server is the lighter path. Minimal CPU is enough to host a persistent spatial map of a venue, and rewards flow when devices query the domain. The same closed-beta gating applies; the same reputation stake requirement applies.

Building on the posemesh is permissionless. Fork the core Rust repo (MIT), use the published Unity SDKSDKSoftware Development Kit. A collection of code libraries, documentation, and tools that lets developers integrate a service into their applications without writing everything from scratch. SDKs are how projects become easy to build with.Like a plug-and-play kit for building furniture. You don't have to mill your own wood, forge your own screws, or design the joinery from scratch. The kit gives you pre-cut parts and instructions so you can assemble the thing in an afternoon.SDK in the glossary → or React Native bindings, and apply for a developer grant of up to $100,000 in AUKI per project. Auki’s 11 September update adds a revamped Auki SDK in a separate MIT-licensed repository (aukilabs/auki-sdk), which Auki calls experimental, while the older Posemesh SDK is the one it runs with customers.

Using Cactus as a retailer requires a paid pilot agreement with Auki Labs at $10,000-plus per chain. The product targets enterprise scale: the Coop Stora Visby reference case is a 5,000-plus-square-metre flagship with 24,000-plus SKUs. Smaller retailers are not the addressable market.

Passive AUKI holding earns nothing. There is no stakingStakingLocking up a cryptocurrency to help secure a blockchain network, usually in exchange for rewards. The locked tokens act as a security deposit that can be taken away if the staker misbehaves.Like putting down a large rental deposit for an apartment. You get the money back if you behave, you earn interest while it's locked, and the landlord takes it if you trash the place.Staking in the glossary → yield, no governance vote, no dividend. The reputation-stake mechanism is reserved for active operators.

Honest assessment

What works. Cactus has at least one externally documented productivity outcome (Coop Stora Visby, ESM Magazine wrap, 15 minutes per user per day saved). The posemesh protocol code is MIT-licensed, though 16 of 37 non-fork repositories carry no licence file, and there is an OpenZeppelin audit clean of high-severity findings. The whitepaper vesting schedule is disciplined (daily linear, 84-month Foundation vest), though code enforces almost none of it. Cross-chain liquidity to peaq is concrete and gives Auki a foothold in a machine-economy DEX network that already exists. Founding-membership of the Intercognitive Foundation with peaq, Geodnet, Mawari and Tashi is the closest thing to a physical-AI consortium in DeAIDeAIDecentralised AI. An umbrella term for blockchain-based projects that build AI infrastructure (compute, data, inference, models, agents) without a single central provider controlling the system.Like the difference between streaming a movie from Netflix and sharing it via BitTorrent. Netflix is fast and polished but one company controls what you can watch and what you pay. BitTorrent is messier but no single operator can shut you out.DeAI in the glossary →; Openmind sits adjacent on the runtime side.

What’s hype. The 40,000-retail-stores figure that surfaced in DeAI commentary is not in any Auki publication. Auki’s own number is 6,500-plus combined chain-footprint with the explicit caveat that the locations are not simultaneously active. The “decentralised reconstruction network” framing describes a network that is in closed beta with a handful of operators. The deflationary supply mechanism is mechanically sound but currently irrelevant: DeFiLlama’s trailing 30 days of on-chain fees annualise to roughly $13,000 . The self-reported “$10 to $20 million ARR by summer” projection from the 5 February 2026 community update is not reflected in DeFiLlama’s revenue line. The Reitan REMA 1000 “1,900 locations” figure is footprint potential per the primary source, not deployment.

Key risks. Centralisation is admitted by both OpenZeppelin and Auki Labs. Cactus revenue is USD-billed, not AUKI-billed; the headline product does not currently drive token demand. The UUPS upgradeable proxy and owner-controlled pause sit on a token that is supposed to settle retail commerce. The 90% concentration of posemesh-core commits across three contributors is consistent with the team size but a soft bus-factorSybilA single entity controlling many wallets that pretend to be independent participants. Used to game airdrops, governance votes, fair-launch allocations, and any system that distributes value or weight per address.Like one person showing up to a free-sample stand wearing twenty different disguises. The booth attendant thinks they are handing out twenty samples to twenty customers; the warehouse runs out twenty times faster, and most of the giveaway ends up in the same kitchen.Sybil in the glossary → flag. The Discord vanity-URL hijack in February 2026 indicates immature security operations for an enterprise project handling retail data. Twenty-four-hour volume is thin against the market cap and there is no top-tier centralised-exchange listing.

Position. Auki sits in the same shelf as peaq on the DePIN side and Openmind on the physical-AI side. Peaq has the broader machine-economy footprint and is the canonical home for AUKI’s cross-chain liquidity. Openmind ships an open-source robot runtime on shipping OEM hardware but its FABRIC coordination layer is mostly forward-dated; AUKI’s posemesh is the spatial-perception counterpart, equally early on the decentralisation curve. Both pair naturally with Geodnet as the positioning piece, which is the framing the Intercognitive Foundation makes explicit. The compute orchestration layer underneath the posemesh is NuNet, which sits on the same architectural-honesty axis as Auki: both publish small but verifiable production figures rather than inflated ones.

Fact: Cactus is billed in USD; AUKI burns are the protocol-layer settlement, not the customer-facing currency. Take: That is honest engineering and the central problem for the token thesis. The retail product does not currently drive AUKI demand. The token earns its place only through the eventual reconstruction-node operator economy and protocol-fee burns from developers and integrators. Neither flow is meaningful at research date. Bet on Auki the company if you bet at all; the AUKI token thesis needs the burn-mint loop to grow by an order of magnitude before the supply mechanic compounds against demand.

Freedom Score: 49/100

D-grade. Disciplined tokenomics and honest open-source posture undercut by admitted protocol centralisation, closed-beta operator network, and an upgradeable token contract with owner-controlled pause.

Dimension	Score	Reasoning
Infrastructure Decentralisation	7/20	OpenZeppelin’s audit explicitly calls the posemesh “currently a centralised system” with MPC wallets controlling privileged on-chain roles. Reconstruction-node network is in closed beta with a handful of operators. No publicly verifiable operator count or geographic distribution.
Governance Decentralisation	4/20	AUKI is utility-only per the whitepaper. No on-chain DAODAODecentralised Autonomous Organisation. A way to coordinate decisions and manage a treasury using token-weighted voting instead of a traditional company structure. Token holders propose and vote on changes directly.Like a shareholder-run company where every shareholder can vote on every decision, the votes are public, and the company can't do anything the shareholders don't approve. The coordination is messier than a normal company but nobody has unilateral control.DAO in the glossary → governance. Posemesh Foundation and Auki Labs Limited (Hong Kong) jointly steer. The audit’s centralisation language applies here too.
Token Distribution Fairness	9/15	Daily linear vesting across each allocation. Posemesh Foundation on an 84-month vest is the longest in the schedule. No day-one major TGE release. Insider weight in the 25-to-35% range is moderate. Community-distributed share under 1% caps the dimension below the A-band.
Censorship Resistance	8/15	UUPS upgradeable proxy on Base means Auki Labs can change the implementation. Pause is owner-controlled and per the audit did not originally pause allowance changes (since fixed). Spatial data is stored off-chain in operator-controlled domain servers, which gives the perception layer reasonable distribution but leaves it exposed during the centralised phase.
Data Sovereignty	10/15	The whitepaper makes privacy-preservation a first-order design goal. Domain servers are operated by the venue owner; spatial data is exchanged peer-to-peer in voluntary clusters. Cactus uses cloud LLMLLMLarge Language Model. A neural network trained on vast amounts of text to predict the next word in a sequence. Modern LLMs (GPT, Claude, Llama, Qwen, DeepSeek) generate human-quality text and are the foundation of most modern AI products.Like an autocomplete that read every book ever written. It has no memory of individual texts but it has absorbed the patterns of language so deeply that it can generate paragraphs that sound human. The skill is statistical, not conscious.LLM in the glossary → providers (OpenAI, Anthropic, Google) as default reasoning, which routes user-facing interaction through centralised third parties and caps the dimension.
Open Source Transparency	11/15	MIT licence on the posemesh core, reconstruction-server and splatter-server, though 16 of 37 non-fork repositories carry no licence file. Contracts source-verified on BaseScan. OpenZeppelin audit published in full. Whitepaper allocation table published. 90% of posemesh-core commits from three contributors, and the live posemesh telemetry dashboard did not return numerical values on plain fetch, both of which pull this back from the upper band.

Returns Score: 34/100

F-grade. Clean utility design and a disciplined vesting schedule on paper that code doesn’t enforce, but value accrual is small enough to be irrelevant at current throughput, Cactus revenue does not drive token demand, and liquidity is thin without a top-tier CEX listing.

Dimension	Score	Reasoning
Token Utility	9/20	Burn-for-credit on Cactus service consumption, operator reputation stake on reconstruction and domain nodes, developer-grant currency. No staking yield, no governance vote, no dividend for passive holders.
Value Accrual	8/20	Each burn triggers a deflationary mint smaller than the burn, trending supply toward a five-billion equilibrium. The mechanism functions but on-chain flow is small (DeFiLlama’s trailing-year fees are under a third of the ~$371,000 annualised incentives it measured in June 2026, so net on-chain earnings are deeply negative). Cactus is billed in USD, not AUKI, which decouples the headline product from token demand.
Supply Dynamics	9/20	The whitepaper sets daily linear vesting, with an 84-month Posemesh Foundation vest. Code enforces little of it: 79.74% of supply sits in deployer-traced plain wallets and only 3.01% in the MerkleVester. The mint is uncapped behind a 4-of-6 Safe, and the top 10 hold 83.8% (24 September 2026 read).
Revenue Sustainability	3/25	DeFiLlama’s trailing-year on-chain fees are small and falling, under a third of the roughly $371,000 of annualised incentive emissions measured in June 2026, so net on-chain earnings are deeply negative. The self-reported “$10 to $20 million ARR by summer” projection from the 5 February 2026 community update is not yet reflected in any independent data source.
Liquidity & Access	5/15	Thin daily volume against the market cap. MEXC is the only centralised venue, with no top-tier listing. The rest of the liquidity sits in Base DEX pools (Uniswap, PancakeSwap, Aerodrome) plus a bridged AUKI/PEAQ pair on MachineX. A mid-size holder rotating would move the price.

Score change log

Date	Score	Change	Reason
2026-10-20	Data	N/A	On-chain fees refreshed: $79,488 trailing year, $254,298 since launch (DeFiLlama, October 2026); net on-chain earnings about negative $292,000. Scores unchanged.
2026-10-04	Data	N/A	Added Auki’s self-reported first gas-leak monitoring contract (5 September update, partner unnamed). Scores unchanged.
2026-10-04	Data	N/A	Added the revamped, MIT-licensed Auki SDK repository from the 11 September update, described by Auki as experimental. Scores unchanged.
2026-09-25	Score	Returns 36 → 34 (F held)	Supply Dynamics 11 → 9: 79.74% of supply sits in deployer-traced plain wallets and 3.01% in the MerkleVester, so code enforces little vesting, and the mint is uncapped behind a 4-of-6 Safe. Supply type set to uncapped.
2026-09-24	Data	N/A	Trailing-year on-chain fees restated to $102,983 (DeFiLlama, 24 Sep), the 10B supply bound reworded as an uncoded initial mint, GitHub and liquidity-venue counts refreshed, the ARR projection re-dated to 5 Feb 2026, and Auki’s pilot footprint no longer called verified. Scores unchanged.
2026-09-23	Data	N/A	Licence wording corrected: the protocol code is MIT, but 16 of 37 non-fork repositories carry no licence file. Scores unchanged.
2026-09-23	Data	N/A	Integrity records re-cut: 3 claims now quote Auki’s own pages verbatim and 4 are our own measurements. Scores unchanged.
2026-08-19	Data	N/A	Correction. The Liquidity evidence carried a hardcoded market cap from CoinGecko, which drifts on every price move. Restated structurally; no score change.
2026-08-15	Data	N/A	Distribution resolved (deferred 2026-08-12). The top 10 hold 83.80% of supply, and all eight unlabelled EOAs among them trace by largest inbound to the token deployer, so the concentration is distribution structure rather than market accumulation. Fairness stays 9/15 pending the monthly.
2026-08-13	Data	N/A	Corrected the supply verdict’s claim text, which read “a fixed 10 billion supply (no inflation)” and contradicted both Auki’s whitepaper and this page: the design is an initial 10B mint deflating toward a 5B equilibrium. Verdict unchanged; neither bound is enforced in code. Attribution moved off www.auki.network, which no longer resolves.
2026-08-12	Data	N/A	WS3 on-chain reconciliation (1->4 verdicts). Supply matches ~10B but is not code-fixed: the Base token is a UUPS upgradeable proxy with an uncapped mint() and pausable transfers, admin + upgrade behind a 4-of-6 Safe (supply_dynamics + token-admin established). Open-source confirmed (aukilabs 53 MIT repos; match). Distribution deferred. Scores unchanged.
2026-08-17	Data	N/A	Data refresh. Trailing-year on-chain fees re-read at $146,781, down 11.8% from the $166,400 June anchor; cumulative $246,615. Seven editorial instances updated, net restated to about -$224,000. Score flagged for the monthly.
2026-06-26	Data	N/A	DefiLlama LlamaAI harvest cross-check plus revenue wired to the structured feed, verified independently (CoinGecko + api.llama.fi). Auki is one of the few DeAI projects DefiLlama tracks with real (tiny) on-chain revenue. Refreshed market data: market cap $22.34M (was $34.83M), FDV $50.33M (was $81.85M), MCap/FDV 0.44, circulating ~44% (was 42.55%), price ~97% below the TGE-day high. Reconciled the on-chain revenue against the DefiLlama dailyRevenue feed and refreshed the meta.primary_data snapshot (cumulative fees $242,573, trailing-year $166,400). That resolved the previously inconsistent fee/incentive/net trio: the stale ”-$47,418 net” was arithmetically impossible, now restated as $166,400 trailing-year fees against ~$371,000 of annualised incentive emissions, net about -$205,000. ATH-source nuance retained: CoinGecko records the ATH as $0.0606 (26 Jan 2025), below our cited TGE-day $0.1812 print (annotated as a thin-liquidity spike). Revenue Sustainability stays 3/25 (trivial flywheel regardless); no score change.
2026-05-14	Editorial	N/A	Cross-link backfill following the NuNet review publish (same day). Added NuNet reference to the reconstruction-node operating description (NuNet Appliance handles node deployment per the Posemesh x NuNet partnership) and to the Position paragraph as the compute-orchestration peer on the architectural-honesty axis. No score change.
2026-05-14	Both	N/A	Initial publish. Freedom 49/100 (D), Returns 36/100 (F). Watch items for next review: reconstruction-node operator count moving out of closed beta, DeFiLlama on-chain fee trajectory off the $43K annualised line, top-tier CEX listing, any clawback or update on the Mars Inc deployment count, the Posemesh Foundation jurisdiction disclosure, and progressive-decentralisation milestones referenced in the 17 May 2025 community update. The “40,000 retail stores” figure referenced in DeAI triage on 13 May 2026 is unverified and was corrected to 6,500-plus combined pilot footprint at publish.