Giza

Giza's Madrid-based founding team (Cem Dagdelen, Fran Algaba, Renç Korzay) has shown unusual commercial adaptability, executing a hard pivot from zkML infrastructure to autonomous DeFi agents and shipping a live product, ARMA, in the process. Reaching a token launch and an institutional agent allocation from Re7 Capital reflects a team that converts thesis into deployed product rather than perpetual research. Against EZKL, which stayed a tool, and Modulus, which was absorbed, Giza found a route to a consumer-facing application. The view would weaken if the pivot proves to be narrative-chasing without retention, and strengthens as ARMA's managed volume compounds.

Giza's technical story spans a genuine zkML heritage — the Orion transpiler and provable ML on Starknet — and a current agent stack that uses verifiability to make autonomous strategies trust-minimized and non-custodial. The differentiation is the pairing: most DeFi agents are not provable, and most zkML projects have no agent. The catch is that, in practice, users may weight agent performance over cryptographic verifiability, blunting the technical edge. Versus general zkVMs and pure agent frameworks, Giza is distinctive but must prove verifiability is a feature users actually demand; a flagship case where provability wins business would move this positive.

The GIZA token went live in May 2025 with utility framed around governance, node-operator staking for protocol security, and seeding agent liquidity for new strategies — a coherent design on paper. The open question is whether agent activity like ARMA's translates into durable token value or whether GIZA trades on the reflexive DeFAI narrative. Compared with pre-token peers like EZKL that retain design optionality, Giza must now defend live economics against still-early usage. Lookout would turn positive on evidence that fees or staking demand track real agent volume, and negative if emissions meet thin organic demand.

Giza has more demonstrable traction than most of its verification cohort: ARMA has reportedly optimized tens of millions in stablecoin deposits across 100,000-plus autonomous trades, with Giza agents cited at $40M-plus in cumulative volume and a $500K institutional allocation from Re7 Capital. That is genuine, measurable usage rather than testnet vanity, and the institutional check is a meaningful signal. The caveat is that DeFi-agent volume can be incentive-driven and cyclical, so durability is unproven. The axis moves up on growing, retained TVL under management, and softens if volume proves mercenary.

Giza has raised roughly $8.2M across a 2023 CoinFund-led pre-seed and a 2025 round with Coinbase Ventures' Base Ecosystem Fund and Arrington Capital, a credible crypto-native cap table with real distribution value. CoinFund's repeat backing signals conviction, and the Coinbase/Base relationship is a tangible go-to-market asset for an agent product. The figure is modest against RISC Zero's tens of millions but strong for an agent-stage project. The view would weaken on a markdown or stalled follow-on, but on backer quality and distribution it is a clear strength.

Giza's pivot lands it in the white-hot autonomous-agent and DeFAI narrative, which gave it momentum and a token bid, but that same narrative is crowded and reflexive, shared with Olas, Theoriq and a wave of agent projects. Its verifiable-ML angle is a differentiator within that crowd, yet the market often rewards agent performance and yield over provability. Against EZKL's quieter, more durable verification framing, Giza traded a defensible niche for a louder, more contested one. The narrative strengthens if 'verifiable agents' becomes a recognized category Giza owns, and stays neutral while it is one DeFAI name among many.

Giza's principal risk is category, not capability: by pivoting into DeFAI it entered crypto-AI's most hype-driven, drawdown-prone corner, where token value can decouple from product and incentive-farmed volume can evaporate. Smart-contract and agent-strategy risk is also real, since autonomous capital allocation can fail in ways static protocols do not. Relative to dormant Modulus or tool-only EZKL, Giza carries live market and execution risk but also live upside. Lookout would de-risk on durable, retained agent TVL and clear token value-capture, and flag it upward if the DeFAI narrative cools while usage proves to have been incentive-bought.