# Giza Review

> Madrid-based lab that ran the ARMA and Pulse DeFi agents on Base. Both were retired in March 2026 and user funds returned; the successor unified agent has no verifiable on-chain AUA or fees. LuminAIR ZKML still roadmap.

- Scores: Freedom 4.2/10 (grade D), Returns 2.8/10 (grade F)
- Category: agent · Token: GIZA
- Published: 2026-05-14 · Updated: 2026-10-04
- Canonical: https://ownyourmind.ai/projects/giza/
- Data (scores + integrity verdicts, JSON): https://ownyourmind.ai/api/projects/giza.json
- Cite as: Own Your Mind (ownyourmind.ai), licence CC BY 4.0. Link the canonical URL.

Charts and interactive widgets are omitted here; their data is in the JSON above.

---

## What it does

Giza is a Madrid-based lab. Through 2025 it ran two autonomous on-chain agents for DeFi yield on Base, ARMA and Pulse. On 26 February 2026 it announced both would be wound down, set a 26 March 2026 migration deadline, returned user funds to their wallets, and folded the line into a single unified agent it markets as Giza World. LuminAIR, a zero-knowledge ML stack co-developed with Starkware's prover team, remains the long-term differentiator and is still pre-production.

The agent model, carried over from ARMA into the unified agent, is non-custodial. A user deploys a smart-account agent that holds their capital and runs an optimisation strategy across DeFi lending protocols such as Aave, Morpho, Compound, and Moonwell, with the optimiser running off-chain on Giza's hosted service and execution landing on Base via account abstraction. Re7 Capital, an institutional trading firm, deployed a 500,000 USDC pilot at GIZA TGE in late May 2025. Those ARMA-era positions were among the funds returned to wallets in the March 2026 wind-down.

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 Coinbase Ventures joining via its Base fund. The Starkware investment dates back to seed; the LuminAIR S-two engineering partnership came later.

## The product in practice

Grounded in our Giza review, the project's on-chain data, and community feedback in Giza's own Discord. The rated product was retired in March 2026, so there is no live agent for us to run today; the hands-on notes below describe the ARMA-era product and are attributed.

**Who it's for.** Stablecoin holders who wanted hands-off DeFi yield without picking pools or timing rebalances themselves. The pitch was a non-custodial smart-account agent that optimised lending positions across Aave, Morpho, Compound, and Moonwell on Base, so a user kept custody while the agent did the work. That suited passive depositors more than DeFi power users who prefer to manage positions themselves.

**What it was like to use (ARMA era).** Through 2025 the community rated ARMA well, with a recurring note in the project's Discord that the agent handled stablecoin yield reliably. The friction was operational: users reported having to stop the agent entirely to withdraw, with no partial-withdrawal option, which made pulling out a few hundred dollars awkward. Others logged a reproducible jitter bug in the personalisation menu when updating an agent, mobile documentation returning 404s, and repeated requests to support more stablecoins and chains. The community also flagged a visibly stretched team, asking for a dedicated community manager to keep pace.

**What it's like now.** The rated agents, ARMA and Pulse, were retired in March 2026 and user funds returned, and the wind-down itself was handled carefully: the co-founder's migration notice shows funds returned to wallets with no deactivation fees, and Pulse positions returned in-kind rather than force-swapped to spare users slippage. The replacement is thin. As of 4 October 2026 the successor Giza World app still loads a data snapshot dated 23 March 2026, the marketing counters on the homepage are fixed values, and on-chain agent balances read near zero. Community engagement wound down with the product, the Discord feedback channel now archived, support ticket-only, and independent Reddit discussion negligible, so there is no live product surface to assess first-hand.

**Demand, honestly.** On the evidence we can verify, demand has collapsed from the ARMA peak. DeFiLlama's adapter read roughly $12,800 of agent positions on 4 October 2026, down from the roughly $4.85 million this review first recorded,  and the headline assets-under-agent figures are a frozen snapshot rather than a measured position. The one part of the protocol demonstrably live on-chain is GIZA staking, a small token economy separate from any agent revenue. Read the marketing AUA and volume figures as historical rather than current.

**The vision.** Giza's real differentiator is verifiable AI: LuminAIR, a zero-knowledge ML proving stack built with Starkware's prover team, would let anyone check on-chain that an agent ran the model it claimed. That is the piece that would make an autonomous agent trustless rather than trusted. It also sits at least a year out, the repository has been dormant since September 2025, and the retired ARMA agent ran on a hosted optimiser with no on-chain verifier. Hold the ambition and the shipping reality separately.

## From two agents to one

For most of 2025 ARMA carried the project. A developer deployed it via Giza's TypeScript SDK, deposited stablecoins or yield-bearing assets, set a strategy band, and the agent re-allocated between lending protocols when yield differentials crossed a threshold. Execution was on-chain on Base, and the intelligence layer that decides when to rebalance ran off-chain on Giza-hosted endpoints. Pulse ran a narrower mandate, automating staked-ETH deployment into Pendle markets.

When this review first published in May 2026 it recorded a DeFiLlama TVL of $4.85 million and a small annualised fee figure for that ARMA product. Both agents had in fact already been retired. Giza announced the wind-down on 26 February 2026, set a 26 March 2026 migration deadline, and returned user funds to their wallets. The figures this review first cited described a product that was being switched off as it published.

The successor is a single unified agent, marketed as Giza World, that inherits the same non-custodial smart-account model on Base. Its public surfaces don't show a current, verifiable position. The gizatech.xyz homepage is a static marketing page: when we checked in June 2026 its headline counters, around $19 million assets under agent and billions in cumulative volume, animated on load but were fixed values, and by 23 September 2026 the assets-under-agent counter rendered $0 in a browser.  The Giza World app loads a frozen data snapshot whose own manifest is dated 23 March 2026, three days before the migration deadline, so its visualisation replays the ARMA-era wind-down rather than live activity.

On-chain, the picture is the opposite of the dashboard. DeFiLlama's adapter sums the USDC positions of Giza smart accounts across Base, Arbitrum, Plasma, and HyperEVM, and on 24 June 2026 it reads about $14,000, down from the $4.85 million this review first recorded. The account-enumeration endpoints the adapter depends on now return errors, so even that residual undercounts rather than measures the live agent. The one part of the protocol that is demonstrably live on-chain is GIZA staking: roughly 46 million GIZA sat in the Staker contract on Base on 24 June 2026 and about 38 million by 24 September 2026, a small but measurable token economy separate from any agent revenue.

## The verifiability pillar (still roadmap)

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

LuminAIR is a Rust library that takes a computational graph from Luminal (a Tinygrad-style ML compiler) 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 substantive engineering work on a confirmed 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 token 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, with roughly a third of it circulating and the rest locked or undistributed. The FDV-to-MCap ratio is roughly 3.0x.

Allocation per Giza's own token-distribution docs (docs.gizaprotocol.ai/token/distribution):

- Early Backers and Investors 31.44%, 36-month linear vesting 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
- Partnerships 6.00%, 3.79% released at TGE

Insider share (team plus investors) is approximately 49.7%. The 36-month linear schedule started at TGE on 29 May 2025, so by September 2026 a little under half of the insider tranche has vested. The remainder grinds out smoothly without cliff shocks.

Token utility, per Giza's listing materials. Operators stake 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, and the fee share is currently paying out nothing. On-chain protocol revenue reads zero across the trailing 24 hours, 7 days and 30 days when re-checked on 18 September 2026, against $127,490 all-time and $56,685 in the trailing year, so the split exists and there is nothing flowing through it. Staked GIZA has fallen about 17% over the same period, to 38.4 million. The other two (governance, gating) are marketing copy without operating evidence. There is no public governance forum, no published DAO 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 that handled institutional capital from Re7 Capital and routing across three chains, that is a meaningful gap.

## How to participate

The friction-free path is the unified agent at world.gizatech.xyz, the successor to ARMA. Connect a wallet, deposit stablecoins, choose a yield strategy, and the agent runs. No GIZA token holding is required, though confirm current status before depositing given the March 2026 reset.

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 hasn't been updated since 28 June 2024. Treat it as deprecated.

## Honest assessment

**What works.** The team is doxxed with relevant mechanism-design and product backgrounds. The Starkware partnership is a substantive engineering relationship at the proving-system level. GIZA staking is demonstrably live on-chain, though the staked balance fell from about 46 million GIZA in June to about 38 million by 24 September 2026.

**What is hype.** The "verifiable AI in production" framing remains marketing, since the agent runs on a hosted optimiser and the on-chain Cairo verifier is still roadmap. The bigger problem now is the activity story. The gizatech.xyz homepage and the Giza World app present large assets-under-agent and volume figures as if live, but the homepage is a static page and the app replays a data snapshot dated 23 March 2026. On-chain measurement of the agent positions reads near zero as of 24 June 2026.

**Key risks.** The rated product was switched off. ARMA and Pulse were retired in March 2026 and user funds returned, and the successor agent has no third-party-verifiable AUA or fee base yet. The single-engineer dependency on LuminAIR is unchanged, with no commits since 3 September 2025. The absence of public audits and the roughly 67% supply overhang both still apply.

**Position.** This sits in the same shelf as Ora Protocol: substantive engineering, an ERC-standards or partnership pedigree, and a wide gap between the technical vision and the shipping product. Ora is the further-gone case, and a useful warning. Its opML oracle did reach production where Giza's ZKML has not, then stopped serving in September 2025, and the team has since repointed the name at an unrelated trading product. Giza reset its agent in March 2026 and is still building the same thing.

## 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.

| Dimension | Score | Reasoning |
| - | - | - |
| Infrastructure Decentralisation | 10/20 | Smart-account execution on Base, multi-chain via LayerZero, but the intelligence layer runs on Giza-hosted IaaS. On-chain Cairo verifier is roadmap. Users trust Giza to run the agreed model. |
| Governance Decentralisation | 4/20 | A Security Council holds executive authority in Phase 1, with community input through off-chain Snapshot votes and forum discussion. No on-chain DAO contracts, no published proposals from non-team participants. |
| Token Distribution Fairness | 4/15 | Insider allocation around 50%. Smooth 36-month linear vest is structurally positive; 2,240 holders on Ethereum indicates thin community float. (figures as of 2026-09-20) |
| Censorship Resistance | 8/15 | Smart-account assets remain user-custodial if Giza pauses its hosted optimiser. The intelligence layer itself is custodial. Meaningful but partial. |
| Data Sovereignty | 8/15 | Funds and strategy parameters stay user-controlled. Model weights and inference live with Giza. ZKML would change this materially; that is roadmap, not state. |
| Open Source and Transparency | 8/15 | `giza-agents` MIT but stale since June 2024. `LuminAIR` has no licence file. `orion` archived. `giza-token` Foundry repo open. No published third-party audits. |

## Returns Score: 28/100

F-grade. The rated revenue base, the ARMA and Pulse agents, was retired in March 2026 and user funds returned. The successor unified agent has no third-party-verifiable AUA or fees, its headline figures are a stale snapshot, and only a small live staking layer remains.

| Dimension | Score | Reasoning |
| - | - | - |
| Token Utility | 9/20 | Operator staking is live on-chain, around 46 million GIZA as of June 2026. Fee share depends on agent fees that are no longer verifiable; governance and gating remain unexercised. |
| Value Accrual | 5/20 | Fee-share-to-stakers mechanism is documented but dormant. With ARMA retired and successor fees unverifiable, no measurable activity links protocol use to token value. |
| Supply Dynamics | 9/20 | Roughly a third circulating. 36-month linear vest with no cliff. FDV/MCap ratio 3.0x. Continuous vest release, no cliff shocks. A little under half of the team-and-investor schedule has elapsed. |
| Revenue Sustainability | 2/25 | ARMA and Pulse retired March 2026, funds returned. Successor AUA and fees unverifiable; DeFiLlama on-chain reads about $14,000 on 24 June 2026. Dashboard figures are a snapshot dated 23 March 2026. |
| Liquidity and Access | 3/15 | Three DEX pools only (Aerodrome Slipstream, DODO on Ethereum, Uniswap V3 on Base). The MEXC, BitMart, Gate.io, KuCoin and XT.COM listings this review named are all gone; Gate returns GIZA as delisted and MEXC has no GIZA pair. Turnover near a tenth of a percent of market cap. |

## Score change log

| Date | Score | Change | Reason |
| - | - | - | - |
| 2026-10-04 | Data | N/A | Product-in-practice sweep: DeFiLlama agent positions refreshed to about $12,800, the Giza World snapshot re-checked as still dated 23 March 2026, and a line implying a live agent corrected to the retired ARMA. Scores unchanged. |
| 2026-10-04 | Data | N/A | Revenue Sustainability evidence now says the trailing-year fee figure is outdated, matching the protocol-revenue verdict (zero fees over 30 days on 18 Sep). Scores unchanged. |
| 2026-09-25 | Data | N/A | Research-data holder count moved from 2,083 (Etherscan, May 2026) to 2,186 (Blockscout, 13 Aug 2026), matching the integrity record. Scores unchanged. |
| 2026-09-24 | Data | N/A | Staked GIZA refreshed to about 38M, liquidity and CEX pro withdrawn, homepage counter now $0, allocation re-attributed to Giza's docs, vest elapsed updated to just under half, holder count bound. Scores unchanged. |
| 2026-09-23 | Data | N/A | Corrected our token record: GIZA's mint is capped in code at 1B and fully issued, so the owner Safe can't add supply; the 2026-08-13 and 2026-09-07 rows said it could. Governance evidence now reflects Giza's published Security Council model; 4/20 held, since no on-chain governance exists. |
| 2026-09-23 | Data | N/A | Integrity records re-cut: 4 claims now quote Giza's own pages verbatim and 5 are our own measurements. Scores unchanged. |
| 2026-09-18 | Data | N/A | Protocol-revenue verdict moved established → outdated: on-chain fees read zero across 24h, 7d and 30d, against $56,685 in the trailing year. Staked GIZA down ~17% to 38.4M; scores flagged for the monthly review. |
| 2026-09-07 | Data | N/A | WS3b censorship rail: GIZA is owned by a 2-of-4 Safe that also holds mint, with no pause and no proxy. Two signatures can issue supply; transfers cannot be halted. Censorship Resistance 8/15 held. |
| 2026-09-07 | Returns | Liquidity & Access 9 → 3 (Returns 34 → 28, F held) | Applied out of cycle. Every CEX listing is gone and daily volume is a three-figure number against a market cap in the low hundreds of thousands, which is the published 0-3 illiquid band rather than the one-or-two-mid-tier-CEXs band the 9/15 described. |
| 2026-09-06 | Data | N/A | WS3b liquidity rail: all five named CEX listings have gone. Gate returns currency GIZA is delisted and MEXC has no GIZAUSDT (control NOSUSDT resolves); the ticker set is three DEX pools with a 0.15% 7-day median turnover. Good-breadth-for-the-cap withdrawn from both surfaces. Liquidity & Access 9/15 flagged for October. |
| 2026-08-13 | Data | N/A | Corrected a verdict authored the same day. We had marked Giza's governance claim inflated against "holders govern through token-weighted voting", a claim Giza does not make and explicitly rejects as "the illusion of direct democracy". Restated to their published council model; verdict now match. The 2-of-4 Safe with uncapped mint is unchanged and still the substance. |
| 2026-08-13 | Data | N/A | On-chain check of the token contract: GIZA is Ownable by a 2-of-4 Safe that holds an uncapped mint(uint256) with no timelock, which is the hard evidence behind the 4/20 Governance score and means the 1B supply is bounded by restraint, not code. Migrated Giza onto the verdict spine. |
| 2026-08-13 | Data | N/A | Holder concentration reads 97.72% top-10 raw but 8.92% once the LayerZero bridge adapter is excluded, across 2,186 holders. GitHub development stopped in March 2026, matching the date the agent dashboard froze. |
| 2026-07-06 | Editorial | N/A | Added a "The product in practice" section (per `specs/product-comparison-framework.md`); the honest read is that the rated product is switched off (ARMA/Pulse retired, dashboards stale, community wound down). No score change. |
| 2026-06-24 | Returns | 44 to 34 (D to F) | Data + re-score. Verified via Giza's X, the arma.xyz redirect, CryptoBriefing, the Giza World data manifest, the DeFiLlama adapter source, and a Base RPC read: ARMA and Pulse were retired (announced 26 Feb 2026, migration deadline 26 Mar 2026) and user funds returned. The successor unified agent has no third-party-verifiable AUA or fees; gizatech.xyz counters are static, the app replays a snapshot dated 23 Mar 2026, and DeFiLlama on-chain reads about $14,000. GIZA staking is live (about 46M GIZA in the Staker on Base). Revenue Sustainability 6 to 2, Value Accrual 9 to 5, Token Utility 11 to 9. Freedom held at 42 (D); the rated architecture is now the unified agent, flagged for the 1 July monthly. Downstream podcast and YouTube assets narrate the retired ARMA fees and need regen. |
| 2026-05-14 | Both | N/A | Initial publish. Freedom 42/100 (D), Returns 44/100 (D). Watch items for next review: LuminAIR commit cadence (currently zero in over eight months), Cairo verifier deployment, third-party audit publication, ARMA fee trajectory off the $38-per-week trough, and any update on the CEO transition flagged in secondary sources. |