Written by: Zack Pokorny

Translated by: Chopper, Foresight News

The landing of AI agents on the blockchain is not smooth. Although blockchain has programmable and permissionless characteristics, it lacks the semantic abstraction and collaborative layer to adapt to agents. The research organization Galaxy has released a report indicating that agents face four major structural frictions on-chain: opportunity discovery, trustworthy verification, data reading, and execution processes. The existing infrastructure is still designed around human interactions, making it difficult to support AI's autonomous management of assets and execution of strategies, which become the core bottlenecks for agents' large-scale landing on the blockchain. The following is a full translation of the report:

The application scenarios and capabilities of AI agents have begun to evolve. They are starting to autonomously execute tasks and are being developed for holding and configuring capital, as well as discovering trading and yield strategies. Although this experimental transformation is still in its very early stages, it is fundamentally different from the previous development model of agents primarily as social and analytical tools.

Blockchain is becoming a natural testing ground for this evolutionary process. Blockchain is permissionless, composable, has an open-source application ecosystem, provides equal access to data for all participants, and all assets on-chain are programmably default.

This raises a structural question: If blockchain is programmable and permissionless, why do autonomous agents still face frictions? The answer does not lie in whether execution is feasible, but in how much semantic and coordination burden exists above execution. Blockchain guarantees the correctness of state transitions but typically does not provide protocol-native abstractions, such as those used for economic explanations, normative identities, or goal-level coordination.

Some frictions stem from architectural defects of permissionless systems, while others reflect the current state of tools, content management, and market infrastructure. In fact, many upper-level functions still rely on software and workflows, which require human participation in their construction.

Blockchain Architecture and AI Agents

The design of blockchain revolves around consensus and deterministic execution rather than semantic interpretation. It exposes low-level primitives such as storage slots, event logs, and call traces, rather than standardized economic objects. Consequently, abstract concepts like positions, yield rates, health coefficients, and liquidity depth usually need to be reconstructed off-chain by indexers, data analysis layers, front-end interfaces, and application programming interfaces, converting the unique states of different protocols into more usable forms.

Many mainstream decentralized finance operational processes, especially those aimed at retail and subjective decision-making, still revolve around users interacting through front-end interfaces and signing individual transactions. This user interface-centered model has expanded with the popularity of retail users, even though a significant part of the activity on-chain is driven by machines. The current mainstream retail interaction model remains: intention → user interface → transaction → confirmation. Programmatic operations follow another path but also face their own limitations: developers select contracts and asset sets during the construction phase, then run algorithms within this fixed scope. Both models fail to accommodate systems that must dynamically discover, assess, and combine operations based on constantly changing objectives at runtime.

When a set of infrastructure optimized for transaction validation is used by systems that need to simultaneously interpret economic states, assess credit, and optimize behaviors around clear goals, frictions begin to appear. This gap partly stems from the permissionless and heterogeneous design characteristics of blockchain and partly arises from interaction tools still built around manual review and front-end intermediaries.

Comparison of Agent Behavior Processes and Traditional Algorithm Strategies

Before discussing the gap between blockchain infrastructure and agent systems, it is necessary to clarify: what is the difference between behavior processes with greater intelligent autonomy and traditional on-chain algorithm systems?

The difference between the two does not lie in the degree of automation, complexity, parameterization settings, or even the ability to dynamically adapt. Traditional algorithm systems can achieve high levels of parameterization, automatically discover new contracts and tokens, allocate funds among various strategy types, and rebalance based on performance. The real distinction lies in whether the system can handle scenarios that were unforeseen during the construction phase.

No matter how complex traditional algorithm systems are, they will only execute preset logic for preconceived patterns. They need to be equipped with pre-defined interface parsers for each type of protocol, predefined evaluation logic that maps contract states to economic meanings, clear credit and normative judgment rules, and hard-coded rules set for each decision branch. When situations arise that do not fit preset patterns, the system either skips them or outright fails. It cannot reason about unfamiliar scenarios; it can only judge whether the current situation matches known templates.

Like this "Digesting Duck" mechanical automatic device, which can imitate biological behavior, but all actions are pre-written.

A traditional algorithm scanning the DeFi lending market can identify familiar events or new deployed contracts matching known factory patterns. But if a new lending infrastructure component with an unfamiliar interface appears, the system cannot evaluate it. A human must check the contract, understand its operational mechanism, determine whether it falls within exploitable opportunities, and write integration logic. Only after this can the algorithm interact with it. Humans interpret, algorithms execute. Agent systems based on foundational models change this boundary. They can achieve:

• Interpretation of vague or incomplete goals. Instructions like "maximize yield while avoiding excessive risk" require semantic interpretation. What counts as excessive risk? How should yield and risk be weighed? Traditional algorithms need to define these conditions precisely in advance, whereas agents can interpret intentions, make judgments, and optimize their understanding based on feedback.
• Generalization to adapt to unfamiliar interfaces. Agents can read unfamiliar contract code, parse documents, or look at application binary interfaces they have never encountered, inferring the economic functions of the system. They do not need to build parsers for each type of protocol in advance. Although this capability is not yet perfect and agents may misjudge what they see, they can attempt to interact with systems that were not anticipated during the construction phase.
• Reasoning under uncertainty in trust and normativity. When credit signals are vague or incomplete, foundational models can probabilistically weigh signals rather than simply applying binary rules. Does this smart contract possess standardization? Based on existing evidence, is the token legitimate? Traditional algorithms either have rules to follow or nothing they can do, while agents can reason about confidence.
• Interpreting errors and making adjustments. When unexpected situations occur, agents can reason about the root of the problem and decide on a course of action. In contrast, traditional algorithms only execute anomaly capture modules, forwarding error information without interpretation.

These capabilities currently exist but are not perfect. Foundational models may produce hallucinations, misjudge content, and make seemingly confident erroneous decisions. In adversarial environments involving capital (where the code can control or receive assets), "attempting to interact with unforeseen systems" may mean financial loss. The core argument of this article is not that agents can reliably execute these functions today, but rather that they can attempt what traditional systems cannot and that future infrastructures will enable these attempts to be safer and more reliable.

This distinction should be viewed as a continuous state rather than an absolute categorical boundary. Some traditional systems may incorporate forms of learned reasoning, while some agents may still rely on hard-coded rules in critical paths. This distinction is directional rather than absolutely binary. Agent systems transfer more interpretation, assessment, and adaptive work to runtime reasoning instead of preset rules during the construction phase. This is crucial for discussing friction issues, as agent systems attempt what traditional algorithms completely avoid. Traditional algorithms avoid discovering frictions by having humans select contract sets during the construction phase; they evade control layer frictions by relying on whitelists maintained by operators; they avoid data frictions by using pre-built parsers for known protocols; and they operate within predetermined safety boundaries to avoid execution frictions. Humans complete work on semantics, credit, and strategy in advance, while algorithms execute within delineated scopes. Early on-chain agent behavior processes might follow this model, but the core value of agents lies in shifting discovery, credit, and strategy assessment to runtime reasoning rather than presumption at the construction phase.

They will try to discover and assess unfamiliar opportunities, reason about standardization without hard-coding rules, interpret heterogeneous states without preset parsers, and execute strategy constraints for potentially vague goals. The presence of frictions does not arise because agents are doing the same things as algorithms but with higher difficulty; rather, it is because they are attempting something entirely different: operating in an open, dynamically interpretable action space rather than a closed, pre-integrated system.

Friction

From a structural perspective, this contradiction does not arise from defects in blockchain consensus but rather stems from the current operation mode of the overall interaction stack developed around it.

Blockchain ensures deterministic state transitions, consensus on the final state, and ultimate determinism. It does not attempt to encode economic meaning interpretations, intention verification, or goal tracking at the protocol layer. These responsibilities have always been borne by front-end interfaces, wallets, indexers, and other off-chain collaborative layers, requiring human intervention.[...]

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。