Giza Review

What it does

Giza is a Madrid-based lab running two intersecting products on the same brand. ARMA is an autonomous on-chain agent for DeFi yield, live on Base since 29 January 2025. LuminAIRZKMLZero-Knowledge Machine Learning. A cryptographic technique that lets a prover convince a verifier that a specific model produced a specific output, without revealing the model weights or the input.A sealed-bid auction with a referee. You don't see the bids and you don't see the model that picked the winner, but the referee hands you a tiny receipt that proves the winner was selected by the agreed rules. ZKML is the maths that produces that receipt.Read more → is a zero-knowledgeZKZero Knowledge. A class of cryptographic proofs that let you prove something is true without revealing any of the underlying information. ZK lets a network verify a transaction without seeing the transaction's contents.Like proving you know the password to a safe by demonstrating you can open it, without ever saying the password out loud. The verifier learns that you know the password and nothing more.Read more → ML framework being co-developed with Starkware’s prover team. ARMA is the shipping artefact. LuminAIR is the differentiator. The two have not yet converged in production.

The ARMA model is non-custodial. A user deploys a smart-accountSmart ContractA program stored on a blockchain that runs automatically when its conditions are met. Smart contracts are how blockchains do anything beyond just transferring tokens — DeFi, NFTs, DAOs, and DeAI infrastructure all run on smart contracts.Like a vending machine. You put in the right input and it produces the expected output, no human operator required. The rules are fixed in the machine itself, anyone can use it, and nobody can stop a transaction in the middle.Read more → agent that holds their capital and runs an optimisation strategy across DeFi lending protocols (Aave, Morpho, Compound, Moonwell). The optimiser itself runs off-chain on Giza’s hosted intelligence service. Execution lands on Base via account abstraction. Re7 Capital, an institutional crypto trading firm, deployed a $500,000 USDC pilot at GIZA TGETGEToken Generation Event. The moment a project's token first becomes tradeable. TGE is when vesting clocks usually start, when liquidity hits exchanges, and when public price discovery begins.Like the IPO day for a startup. Everything that happened before TGE was private valuations and paper agreements. Everything after is the public market deciding what the thing is worth in real time.Read more → in late May 2025.

The lab was founded in October 2022 by three doxxed co-founders. Fran Algaba is CEO. Cem Dagdelen previously founded Curve Labs, a mechanism-design studio behind Toucan, API3, and LabDAO. Renç Korzay leads product, with prior PM roles at Sony, EA, and Johnson & Johnson. Total raised across two rounds is $8.2 million. The seed round in July 2023 was led by CoinFund with Starkware and Arrington Capital participating. The Series A in May 2025 was led by CoinFund again, with Base Ecosystem Fund (via Coinbase Ventures) joining. The Starkware investment dates back to seed; the LuminAIR S-two engineering partnership came later.

The shipping artefact

ARMA is the load-bearing product. The smart account abstracts a portfolio agent. A developer deploys it via Giza’s TypeScript 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.Read more →, deposits stablecoins or yield-bearing assets, sets a strategy band, and the agent re-allocates between lending protocols when yield differentials cross a threshold. The execution layer is fully on-chain on Base. The intelligence layer that decides when to rebalance runs off-chain on Giza-hosted endpoints.

DeFiLlama records the protocol’s TVLTVLTotal Value Locked. The sum of all assets currently deposited in a protocol's smart contracts. TVL is the standard measure of how much capital a DeFi or DeAI protocol is custodying.Like the assets under management of a hedge fund. AUM tells you how much money the fund has been trusted with, which is a rough proxy for how much business it's doing. TVL plays the same role for crypto protocols.Read more → at $4.85 million on 14 May 2026. That breaks down as $2.84 million on Arbitrum, $2.00 million on Base, with smaller traces on Plasma and a Hyperliquid L1 footprint. Annualised fees are $194,000. Cumulative fees since launch sit at $125,000. These figures are independently measured. They are also small, and the 7-day fee print of $38 indicates the headline activity has cooled significantly from the launch quarter.

Giza’s own dashboard tells a much larger story. The lab’s communications cite $1.5 billion of cumulative agentic volume, $16 million peak TVL, around 60,000 holders, and 36,000 deployed agents. None of these throughput figures are independently verifiable from on-chain data. Treat them as Giza-reported, the way a SaaS company reports gross transaction volume: indicative of activity, not of the protocol-level fee base that accrues to the 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.Read more →.

The verifiability pillar (still roadmap)

The marketing positions Giza as a verifiable AI company. The reality is two-layered.

LuminAIR is a Rust framework that takes a computational graph from Luminal (a Tinygrad-style ML framework) and compiles it to a Circle STARK proof using Starkware’s S-two prover. The roadmap is three phases. Phase 1 is a set of 11 primitive operators (Sin, Sqrt, Add, Mul) and a working prover. Phase 2 adds fused operators and a Python SDK. Phase 3 deploys a Cairo verifier contract on-chain so anyone can verify a LuminAIR proof from a smart contract. The Starkware partnership announcement in June 2025 stated that Phase 3 was still future work.

Three observations about LuminAIR’s current state, from primary sources:

The repository has had no commits in over eight months. Last push was 3 September 2025.

One engineer, Raphael Doukhan, authored 471 of approximately 547 commits and is the named Starkware partnership lead. The bus factor is one.

LuminAIR has no licence file in the repository. The legal default in that situation is restrictive: any downstream project wanting to fork or build on LuminAIR has to ask permission.

The Rekt News recommender, named in the Starkware announcement as the showcase deployment, has not received a commit to its dedicated repo since 18 June 2025. No on-chain Cairo verifier deployment has been announced.

The original ZKML stack, orion, which ran on StarkNet via the Cairo language, was archived on 3 March 2025. The team has effectively swapped the StarkNet-native runtime for a Starkware-prover-with-Base-execution model, where the verifier piece sits ahead of them.

ARMA does not use LuminAIR proofs in production today. The agent’s strategy decisions run on Giza’s hosted optimiser. No public verifier contract exists that an end-user could query to confirm a particular ARMA rebalance was produced by the agreed model. The verifiable-AI pillar is genuine engineering work on a genuine Starkware partnership; it also sits at least a year out from being part of the shipping product. Hold those two facts separately.

Tokenomics

The GIZA 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.Read more → launched on 29 May 2025 across Ethereum, Base, and Arbitrum as a LayerZero omnichain fungible token. The Ethereum contract 0x590830dFDf9A3F68aFCDdE2694773dEBDF267774 is verified, written in Solidity 0.8.26, and not proxied. The contract name in the source is GizaMainChain.

Total supply is 1 billion GIZA. Circulating supply is approximately 331 million, around 33% of total. The FDVFDVFully Diluted Valuation. The market cap a token would have if every token that will ever exist were already in circulation. FDV is what the project would be worth if all locked, vesting, or unminted tokens were trading today.Like valuing a startup based on what every share would be worth if all the unvested employee options had already been exercised. The number is bigger and uglier than the official market cap, but it tells you the true ceiling.Read more →-to-MCap ratio is roughly 3.0x, with about 67% of supply still locked or undistributed.

Allocation per third-party aggregator data (Giza’s own tokenomics blog post returns a 403 to public fetches; this figure is Tier-2 and worth flagging):

• Early Backers and Investors 31.44%, 36-month linear vestingVestingA schedule that locks up tokens allocated to insiders, investors, and team members, releasing them gradually over months or years. Vesting prevents insiders from dumping on public buyers immediately after launch.Like a new employee's stock options at a startup. You don't get all the shares on day one. They unlock over four years so you stick around and do the work rather than cashing out and leaving.Read more → from TGE, no cliff
• Community Distribution 22.21%, 2.21% released at TGE
• Treasury and Future Distributions 22.10%, 48-month linear, no cliff
• Team and Advisors 18.25%, 36-month linear, no cliff
• Ecosystem Partnerships 6.00%, 3.79% released at TGE

Insider share (team plus investors) is approximately 49.7%. At today’s date, about 11.5 of the 36-month linear schedule has elapsed, so roughly a third of the insider tranche has vested. The remainder grinds out smoothly without cliff shocks.

Token utility, per Giza’s listing materials. Operators stakeStakingLocking 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.Read more → GIZA to back agent liabilities and secure the execution network. A portion of every autonomous transaction fee flows to stakers and the treasury. Token-weighted voting steers protocol parameters, agent onboarding, and revenue-split proposals. Holding the token gates access to the agent launchpad, AI marketplace, and analytics.

Two of those four utility statements are documented mechanics: operator staking and fee share. The other two (governance, gating) are marketing copy without operating evidence. There is no public governance forum, no published 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.Read more → proposals, no on-chain treasury vote that anyone outside the team has cast. The token has utility on paper that exceeds the utility currently exercised.

There is also no public third-party security audit. ARMA’s smart-account contracts, the LayerZero OFT cross-chain plumbing, and the token contract itself all sit without published audit reports. For a protocol handling institutional capital from Re7 Capital and routing across three chains, that is a meaningful gap.

How to participate

The friction-free path is to use ARMA at agent.gizatech.xyz. Connect a wallet, deposit stablecoins, choose a yield strategy, and the agent runs. No GIZA token holding required.

The token-holder path adds operator staking (claimed but with no public operator-economics doc) and governance voting (claimed but with no operating forum) plus access to the agent launchpad.

The developer path uses @gizatech/agent-sdk from npm to deploy custom agents. The Python giza-agents library is the legacy path and has not been updated since 28 June 2024. Treat it as deprecated.

Honest assessment

What works. ARMA is in production. Re7 Capital is a confirmed institutional pilot. DeFiLlama-verified fees of $194,000 annualised are small but measurable from a third-party source. The Starkware partnership is a substantive engineering relationship at the proving-system level. The founding team is doxxed with relevant mechanism-design and product backgrounds. Liquidity is solid for the cap.

What is hype. The “verifiable AI in production” framing. ARMA is autonomous-agent execution on Base with the intelligence running on a hosted optimiser. Verifiability sits on the roadmap, several quarters out from being a user-facing property of ARMA itself. The self-reported $1.5 billion of agentic volume is also marketing. The measurable third-party figure is $194,000 annualised fees on $4.85 million TVL.

Key risks. The single-engineer dependency on LuminAIR is concrete. If Raphael Doukhan stops shipping, the proving stack stops shipping. The absence of public audits matters more as institutional capital grows. The 67% supply overhang releases smoothly but it does release. The original StarkNet-native stack was archived, which is a normal pivot but worth noting: the team has shifted runtime models once before.

Position. This sits in the same shelf as Ora Protocol: substantive engineering, small revenue, an ERC-standards or partnership pedigree, and a wide gap between the technical vision and the shipping product. Ora’s opML is in production where Giza’s ZKML is not. Giza’s autonomous agent product is in production where Ora’s equivalent operates at much smaller scale. Different sides of the same trade-off.

Freedom Score: 42/100

D-grade. The technique decentralises on the execution side and centralises on the intelligence side. The token-holder protections are mostly marketing copy.

Returns Score: 44/100

D-grade. Real shipping fees on a multi-chain liquid token, with a small revenue base, a wide marketing-versus-measurement gap, and a flagship verifiable-AI product still in roadmap state.

Score change log