YZi Labs Backs De¹’s Financial World Model

De¹ is bringing a live-market feedback problem into the design of autonomous financial agents. YZi Labs has backed the company through EASY Residency Season 4, where each of the 24 selected companies received $500K.

The investment was announced for De¹ on August 28, 2026. No valuation, ownership percentage, security type, or priced-round label was disclosed. The most accurate description is a $500K strategic investment through YZi Labs' accelerator and residency program.

De¹ is building what it calls a Financial World Model for autonomous on-chain trading. The premise is that financial agents need more than access to historical patterns and a mechanism for proposing trades. They need an environment in which simulation, reasoning, execution, and the consequences of execution can inform one another.

Why Live Execution Changes the Problem

A conventional backtest holds an important variable still: the strategy's own effect on the market. Historical data can tell an agent what prices and volumes were available. It cannot fully reveal how liquidity, slippage, and the next decision would have changed if that agent had placed a specific order at a specific moment.

That distinction becomes more important as software gains authority to move capital. A recommendation can be evaluated without ever reaching an exchange. An autonomous agent has to live with the position it actually receives, including partial fills, fragmented liquidity, fees, latency, and the way its action changes the next available choice.

De¹ describes its architecture as a loop connecting market simulation, financial reasoning, on-chain execution, and verifiable feedback. The system is intended to let agents train in simulated environments, act across live venues, and carry observed outcomes back into subsequent decisions. The company reports integrations with more than 1,000 liquidity sources across more than 40 chains. Those are De¹'s figures rather than independently audited metrics.

The underlying idea is useful even before the architecture has proved itself at scale. A financial model should not be graded only on whether it predicted a direction. It should also be evaluated on whether it translated that prediction into an executable action, how the market responded, and whether the resulting record improved the next decision.

What YZi Labs Is Backing

YZi Labs selected De¹ for the fourth season of EASY Residency, its program for companies working across Web3, artificial intelligence, and biotechnology. The investor says it reviewed more than 1,500 applications and selected 24 companies, with each receiving $500K. De¹ appears in the cohort's Web3 group.

The capital arrives shortly after De¹'s public launch announcement. That sequence makes the investment less a reward for a mature operating record than support for an early technical thesis: autonomous finance will require a world model grounded in market mechanics and accountable execution.

For YZi Labs, De¹ also fits an investment pattern around infrastructure that can make on-chain systems useful to software agents. An agent that can reason about a market but cannot interact reliably with fragmented liquidity remains a demonstration. An agent that can execute but cannot learn from the result is automation without a durable feedback loop.

De¹ is attempting to join those layers. Its model treats a market as an environment with state, actions, outcomes, and feedback rather than a static sequence of prices. The challenge is to make the environment realistic enough that learning transfers to live markets, while making the execution record clear enough that users can evaluate why an agent behaved as it did.

The On-Chain Context

On-chain markets offer an unusual setting for this work. Transactions produce public, machine-readable records, and assets can move through programmable systems without the same manual handoffs found in traditional finance. At the same time, liquidity is distributed across chains, protocols, pools, and venues. Execution quality depends on routing, timing, fees, smart-contract behavior, and the changing conditions created by other participants.

That makes verifiability necessary but not sufficient. A blockchain can prove that a trade occurred. It does not by itself prove that the trade was sensible, that the agent understood the risk, or that a favorable result came from a repeatable process rather than chance. De¹'s opportunity is to connect the verifiable event to the reasoning and learning loop around it.

If successful, that connection could support more rigorous evaluation of financial agents. Operators could compare intended and realized positions, trace the effect of execution, test whether feedback changed later behavior, and separate attractive explanations from accountable outcomes.

What Remains to Prove

The investment does not settle whether a Financial World Model can remain calibrated across volatile markets and rapidly changing on-chain infrastructure. Simulations can become stale. Feedback can reward short-term outcomes that hide long-term risk. Agents can also amplify one another when many systems learn from similar signals and act through the same liquidity.

De¹ will need to show that its training environments capture the mechanics that matter, that live feedback improves decisions instead of reinforcing noise, and that users can set meaningful limits on autonomous behavior. Reliability will depend on what happens during adverse conditions, not only on the quality of a normal-market demonstration.

The company will also have to turn broad connectivity into dependable execution. Access to more venues can improve opportunity and routing, but it adds operational surfaces, smart-contract dependencies, and new failure modes. A useful world model must understand not only where capital can move, but when it should refuse to move at all.

That is the obligation inside this $500K bet. YZi Labs is backing an attempt to make autonomous financial intelligence answerable to the market consequences it creates. The important evidence will arrive when De¹ can show that an agent learned something durable from the difference between the trade it imagined and the fill it actually received.