As DeFi expands in scale and complexity, AI-driven "Agentic Finance" is becoming a key direction to lower the barrier of entry for users. This article systematically analyzes the current development and core challenges of two types of intelligent agents:

Co-pilot Agents
Platforms like &milo, The Hive, and Meridian assist users with investment decisions, asset rebalancing, and more through natural language processing. However, they still face issues such as execution errors, data latency, and clumsy permission management. For now, they primarily serve as "teachers" or guides rather than fully autonomous systems.

Quantitative Agents
Projects like SendAI, Lomen, Unblinked, and Giza’s ARMA agent automate complex strategies, enabling fully autonomous asset management and yield optimization. Yet, they remain vulnerable to real-world challenges like market volatility and protocol pauses, highlighting their operational fragility.

Core challenges include a lack of transparency in agent decision-making ("black box" operations), imperfect permission mechanisms, unrealistic testing environments, and unverifiable execution outcomes.

Solutions Rely on Two Key Foundations
Verifiable Reliability: Technologies like trusted execution environments (TEEs) can ensure agent actions are auditable and traceable, establishing verification standards for different risk levels.
Agent-Friendly Protocols: DeFi protocols must provide stable execution interfaces, secure permission controls, and consistent data structures.

The article emphasizes that DeFi’s next milestone is not about scaling larger but about building trust in automation. Over the next six months, co-pilot agents are expected to improve rapidly. In the long term, AI agents may become the default interaction layer for DeFi, but this depends on robust underlying infrastructure rather than just smarter AI models.

Original Title: DeFi's next milestone: What it'll take for agentic finance to work
Original Author: @Lemniscap
Compiled by: Ismay, BlockBeats

Editor’s Note:
When the world of DeFi becomes too complex even for professional users, how can we return control to ordinary people?

This research article from Lemniscap systematically outlines the rise and challenges of "Agentic Finance." Early products like &milo, Meridian, SendAI, and The Hive demonstrate how AI can become a new interface for on-chain interactions, while also exposing significant gaps in execution reliability, permission security, and verification mechanisms. The author argues that the key to DeFi’s next stage lies not in smarter models but in more trustworthy underlying structures—ensuring every agent action is verifiable, traceable, and reliable.

This is not just a turning point in technological evolution but also an experiment in rebuilding trust. As the article states: DeFi’s next milestone is not greater scale but trust in automation.

By 2025, DeFi looks nothing like its early days.

The data speaks for itself: institutional funds have poured over $10 billion into DeFi in a single quarter, with active protocols spread across dozens of chains exceeding 3,000. The total value locked (TVL) in DeFi protocols reached $160 billion in 2025, a 41% year-on-year increase, while cumulative trading volume on DEXs and perpetual platforms reached "trillions."

As DeFi’s scale and capabilities grow, so does its complexity. Most people simply cannot keep up with everything happening on-chain. If we want more users to seize these new opportunities, we must build tools that help them make better decisions effortlessly—and this is the future direction.

Meanwhile, AI has integrated into daily life, fostering new habits around automation. This trend has given rise to "Agentic Finance"—intelligent agents handling the navigation and execution of financial operations.

Even simple browser-based agents like Comet demonstrate the rapid evolution of such tools. When you execute a DeFi operation through a browser agent (as illustrated by SendAI founder Yash), you glimpse the potential of Agentic Finance.

The vision is intuitive: instead of scouring dashboards or lengthy X threads, you simply tell the AI your goal, and it automatically handles the steps to achieve it.

Two types of agents are emerging:

• Co-pilots: Guide users through DeFi decision-making.

• Quant Agents: Focus on automated strategy execution, akin to "autopilots."

Both are still in early stages with flaws, but they point to a new direction—a fundamentally different, AI-driven way of interacting with DeFi.

Intelligent Agents as "Co-pilots"
Think of these agents as personal assistants. Instead of checking charts or jumping between protocols, you ask questions in natural language, like "What are the hottest tokens right now?" or "Where can I earn the highest yield?" The agent responds directly with answers and next-step suggestions—like a knowledgeable, on-call friend.

For example, &milo’s co-pilot mode assists with investment decisions, asset rebalancing, and portfolio insights—simplifying operations while keeping you in control.

With natural language explanations and smart prompts, &milo helps users understand positions and compare yield opportunities without digging through dashboards. It showcases how co-pilot agents are evolving from simple chat assistants into full-fledged DeFi guides.

To observe these agents in action, we tested several newly launched products, assessing their ability to handle real DeFi tasks.

The results revealed limitations. For instance, an agent could identify popular tokens but failed to execute a purchase. Two transactions failed due to "insufficient balance" errors, even though the account had enough SOL for fees.

The Hive took a different approach—organizing multiple DeFi agents into a "swarm" that collaboratively handles cross-chain operations, yield strategies, and liquidation defense, all coordinated through a clean chat interface. This network of specialized agents can execute multi-step on-chain actions via natural language commands.

We tested the same buy order on The Hive. It identified the popular token WEED but executed the purchase with the wrong contract address.

Overall, Milo demonstrates how to integrate portfolio management into a seamless workflow, while The Hive explores multi-agent collaboration. As agents advance, they are developing clearer specializations.

For example, Meridian targets beginners—helping them take their first steps into DeFi. Its mobile-first design and clear prompts make basic operations like swapping, staking, or checking yields more accessible.

Meridian performs smoothly and executes quickly on core tasks. More importantly, it knows its limits. When users request unsupported actions, it explains why instead of blindly attempting—making it a reliable starting point for newcomers.

Meridian founder Benedict explained:
"Meridian lets users securely research and operate using natural language. We’ve made the agent’s research features free to the public at meridian.app. Mobile app registrants can use the agent’s swap, multi-swap, and portfolio purchase functions. Accounts are still in closed beta; interested users can contact @bqbrady on Twitter to apply."

Our testing shows that most AI agents focused on DeFi navigation currently act as "teachers" or "assistants," primarily helping with basic operations like swapping.

Reliably handling more complex processes—such as providing liquidity or managing leveraged positions—requires further refinement.

As Rishin Sharma, AI Lead at the Solana Foundation, noted:
"Large language models (LLMs) are prone to hallucinations when handling broad tasks and struggle with deterministic operations. Function-calling mechanisms like MCP may be better suited to translating 'action plans' into execution. While LLMs excel at ideation and guidance, they fall short in precise execution. For Agentic Finance to become truly reliable, we must move beyond LLMs to develop specific function-calling mechanisms, clear execution strategies, verifiability, and secure permission systems. In other words, today’s agent execution layer is underdeveloped—the AI 'brain' is smart enough, but it lacks a robust 'body.'"

Intelligent Agents as "Autopilots"
If co-pilot agents are like mentors, quantitative agents are like autopilots. They not only design strategies but also execute them—monitoring markets, testing trades, and acting at machine speed, enabling fully automated DeFi strategies.

A emerging example is SendAI. It is not a quant agent itself but a toolkit for building them. Its "Agent Kit" for Solana supports over 60 autonomous actions, including token swaps, new asset issuance, and lending management, directly interacting with major protocols like Jupiter, Metaplex, and Raydium.

In other words, it provides developers with a "rail system" to connect decision-making models to on-chain execution.

SendAI founder Yash summarized their vision:
"We believe every AI agent will have its own wallet in the future. SendAI is building the tools and economic layer for this ecosystem, enabling agents to perform any action on Solana. We’re creating a platform for context-aware agents that support long-running, persistent, and asynchronous complex tasks."

Meanwhile, other teams are making this capability more accessible. Lomen curates strategies and lets users "deploy with one click," lowering the barrier to quant automation without coding.

For advanced users who prefer custom systems, Unblinked offers an AI-powered strategy experimentation environment. It’s like Cursor for trading: users sketch strategy ideas, run and optimize them in a safe sandbox, and then decide whether to deploy real capital.

Some platforms combine multiple agents to complete tasks.

For instance, Almanak integrates "programming agents" and "backtesting agents": users describe strategies in natural language, the AI generates production-ready code, backtests it with over 10,000 Monte Carlo simulations, and delivers a battle-tested strategy.

Finally, some teams focus on real-time market advantages.

Giza’s ARMA agent dynamically allocates funds across lending protocols to maximize stablecoin yields. Instead of leaving funds in a single pool, ARMA continuously monitors rates, liquidity, and gas costs, moving assets accordingly. Its flagship agent has managed over $17 million, claiming an 83% higher yield than static holdings.

Overall, these quant agents significantly reduce time costs and expose ordinary users to strategies once reserved for professional teams. However, they also reveal the fragility of automation: agents can still "stumble" during data delays, protocol pauses, or market volatility.

In other words, they make you faster but far from "invincible."

The Challenges They Face
After spending time with current agents, you’ll notice recurring issues: they sometimes suggest actions for deprecated pools, rely on outdated data, and fail to adapt when multi-step plans go awry, repeatedly retrying the same failed action.

Permission management is also clumsy—users must either grant full wallet access or manually approve every minor step. Testing is superficial, with simulations failing to replicate real-world chaos like sudden liquidity shifts or governance parameter changes.

One of the most serious issues is that these agents largely operate as "black boxes."

Users can’t see which inputs were read, how options were weighed, whether real-time states were checked, or why a specific transaction was chosen. Without signed operation records, there’s no way to verify if "promised results" match "actual execution."

Users must "babysit" the automation—inefficient and making performance hard to evaluate.

Without mechanisms to verify decisions and prove actions adhere to stated strategies, users can’t distinguish "reliable systems" from "well-marketed hype."

For larger capital, DeFi platforms must shift from "trust us" to "verify for yourself." This is the critical turning point for building "auditable, governable, and trustworthy" Agentic Finance infrastructure.

The Infrastructure Gap
The core issue is the lack of foundational tools to ensure agents are trustworthy, consistent, and secure at scale. To fix this, we need infrastructure that verifies agent behavior, confirms execution results, and enforces uniform rules across all environments. Only then will people confidently entrust real capital.

Most users don’t care about the agent’s "thought process"—they just want confirmed outputs that are correct, verified, and within safe boundaries. For building trust, "verifiable reliability" matters more than "visibility."

This is the essence of "Verifiable Reliability." Agents don’t need to log every internal step but should operate under clear strategies and reasonable checks: spending caps, execution time windows, confirmations for critical actions, etc.

Under the hood, these rules can be enforced via trusted execution environments (TEEs) or similar systems—proving compliance without exposing all details. The result: auditable outputs when needed and operations ordinary users can trust immediately.

This verification layer need not be one-size-fits-all. Everyday use cases can adopt lightweight security and standardized metrics, while high-risk or institutional scenarios can require stronger proofs and formal verification. The key is that each infrastructure layer should provide measurable reliability matching its risk level.

Preparing Protocols for Agents
The next step is making protocols "agent-friendly."

Most DeFi protocols aren’t designed for intelligent agents. They need to provide more stable, secure execution interfaces: operation previews, safe retries, and consistent data structures for execution. Permission designs should be "scoped" rather than "all-or-nothing," letting agents operate within clear boundaries without controlling the entire wallet.

Without these foundations, even the smartest agent frameworks will be tripped up by fragile underlying layers. Once these bases are solid, users won’t need to manually monitor automation; development teams can reduce debugging time and focus on innovation; execution results across service providers will be comparable against shared benchmarks—moving beyond marketing claims.

What Needs to Change
The solution is straightforward: make agents provable and protocols agent-ready. Add a strategy layer between agents and wallets, and require all execution processes to be traceable and verifiable, not "black-box."

For example, Termina’s SVM engine is built on this idea—providing AI agents with a real Solana runtime environment for modeling, decision-making, and learning based on on-chain data. Meanwhile, protocols should open "dry-run" interfaces, clear error codes, safe retry mechanisms, consistent core data structures (positions, fees, health factors), and session-based permission controls.

When these features are implemented, users are freed from "babysitting" agents; teams reduce system failures; institutional investors finally get the safety rails and verifiable proofs they need.

Realistic Timeline
Over the next six months, "co-pilot" agents are expected to improve the fastest. Better data pipelines will enhance their reliability in everyday use.

Within a year, as testing standards strengthen, agents will coordinate execution across protocols, with humans only approving critical steps. Longer-term, as infrastructure matures, intelligent agents may blur into DeFi’s default interaction layer—no longer separate "tools" but the primary way people interact with the financial system daily.

Conclusion
"Agentic Finance" is lowering participation barriers, making automation accessible beyond experts. But to operate at scale, it needs better "foundations": real-time data, safer permission mechanisms, stronger testing systems, and more transparent execution results.

Smarter AI alone won’t solve these problems. Real progress will come from improving underlying structures.

DeFi’s next milestone is not just scaling larger—it’s about trust in automation. And that day will only arrive when AI agents evolve from "proof-of-concept demos" into truly reliable executors.

As DeFi expands in scale and complexity, AI-driven "Agentic Finance" is becoming a key direction to lower the barrier of entry for users. This article systematically analyzes the current development and core challenges of two types of intelligent agents:

Co-pilot Agents
Platforms like &milo, The Hive, and Meridian assist users with investment decisions, asset rebalancing, and more through natural language processing. However, they still face issues such as execution errors, data latency, and clumsy permission management. For now, they primarily serve as "teachers" or guides rather than fully autonomous systems.

Quantitative Agents
Projects like SendAI, Lomen, Unblinked, and Giza’s ARMA agent automate complex strategies, enabling fully autonomous asset management and yield optimization. Yet, they remain vulnerable to real-world challenges like market volatility and protocol pauses, highlighting their operational fragility.

Core challenges include a lack of transparency in agent decision-making ("black box" operations), imperfect permission mechanisms, unrealistic testing environments, and unverifiable execution outcomes.

Solutions Rely on Two Key Foundations
Verifiable Reliability: Technologies like trusted execution environments (TEEs) can ensure agent actions are auditable and traceable, establishing verification standards for different risk levels.
Agent-Friendly Protocols: DeFi protocols must provide stable execution interfaces, secure permission controls, and consistent data structures.

The article emphasizes that DeFi’s next milestone is not about scaling larger but about building trust in automation. Over the next six months, co-pilot agents are expected to improve rapidly. In the long term, AI agents may become the default interaction layer for DeFi, but this depends on robust underlying infrastructure rather than just smarter AI models.

Original Title: DeFi's next milestone: What it'll take for agentic finance to work
Original Author: @Lemniscap
Compiled by: Ismay, BlockBeats

Editor’s Note:
When the world of DeFi becomes too complex even for professional users, how can we return control to ordinary people?

This research article from Lemniscap systematically outlines the rise and challenges of "Agentic Finance." Early products like &milo, Meridian, SendAI, and The Hive demonstrate how AI can become a new interface for on-chain interactions, while also exposing significant gaps in execution reliability, permission security, and verification mechanisms. The author argues that the key to DeFi’s next stage lies not in smarter models but in more trustworthy underlying structures—ensuring every agent action is verifiable, traceable, and reliable.

This is not just a turning point in technological evolution but also an experiment in rebuilding trust. As the article states: DeFi’s next milestone is not greater scale but trust in automation.

By 2025, DeFi looks nothing like its early days.

The data speaks for itself: institutional funds have poured over $10 billion into DeFi in a single quarter, with active protocols spread across dozens of chains exceeding 3,000. The total value locked (TVL) in DeFi protocols reached $160 billion in 2025, a 41% year-on-year increase, while cumulative trading volume on DEXs and perpetual platforms reached "trillions."

As DeFi’s scale and capabilities grow, so does its complexity. Most people simply cannot keep up with everything happening on-chain. If we want more users to seize these new opportunities, we must build tools that help them make better decisions effortlessly—and this is the future direction.

Meanwhile, AI has integrated into daily life, fostering new habits around automation. This trend has given rise to "Agentic Finance"—intelligent agents handling the navigation and execution of financial operations.

Even simple browser-based agents like Comet demonstrate the rapid evolution of such tools. When you execute a DeFi operation through a browser agent (as illustrated by SendAI founder Yash), you glimpse the potential of Agentic Finance.

The vision is intuitive: instead of scouring dashboards or lengthy X threads, you simply tell the AI your goal, and it automatically handles the steps to achieve it.

Two types of agents are emerging:

• Co-pilots: Guide users through DeFi decision-making.

• Quant Agents: Focus on automated strategy execution, akin to "autopilots."

Both are still in early stages with flaws, but they point to a new direction—a fundamentally different, AI-driven way of interacting with DeFi.

Intelligent Agents as "Co-pilots"
Think of these agents as personal assistants. Instead of checking charts or jumping between protocols, you ask questions in natural language, like "What are the hottest tokens right now?" or "Where can I earn the highest yield?" The agent responds directly with answers and next-step suggestions—like a knowledgeable, on-call friend.

For example, &milo’s co-pilot mode assists with investment decisions, asset rebalancing, and portfolio insights—simplifying operations while keeping you in control.

With natural language explanations and smart prompts, &milo helps users understand positions and compare yield opportunities without digging through dashboards. It showcases how co-pilot agents are evolving from simple chat assistants into full-fledged DeFi guides.

To observe these agents in action, we tested several newly launched products, assessing their ability to handle real DeFi tasks.

The results revealed limitations. For instance, an agent could identify popular tokens but failed to execute a purchase. Two transactions failed due to "insufficient balance" errors, even though the account had enough SOL for fees.

The Hive took a different approach—organizing multiple DeFi agents into a "swarm" that collaboratively handles cross-chain operations, yield strategies, and liquidation defense, all coordinated through a clean chat interface. This network of specialized agents can execute multi-step on-chain actions via natural language commands.

We tested the same buy order on The Hive. It identified the popular token WEED but executed the purchase with the wrong contract address.

Overall, Milo demonstrates how to integrate portfolio management into a seamless workflow, while The Hive explores multi-agent collaboration. As agents advance, they are developing clearer specializations.

For example, Meridian targets beginners—helping them take their first steps into DeFi. Its mobile-first design and clear prompts make basic operations like swapping, staking, or checking yields more accessible.

Meridian performs smoothly and executes quickly on core tasks. More importantly, it knows its limits. When users request unsupported actions, it explains why instead of blindly attempting—making it a reliable starting point for newcomers.

Meridian founder Benedict explained:
"Meridian lets users securely research and operate using natural language. We’ve made the agent’s research features free to the public at meridian.app. Mobile app registrants can use the agent’s swap, multi-swap, and portfolio purchase functions. Accounts are still in closed beta; interested users can contact @bqbrady on Twitter to apply."

Our testing shows that most AI agents focused on DeFi navigation currently act as "teachers" or "assistants," primarily helping with basic operations like swapping.

Reliably handling more complex processes—such as providing liquidity or managing leveraged positions—requires further refinement.

As Rishin Sharma, AI Lead at the Solana Foundation, noted:
"Large language models (LLMs) are prone to hallucinations when handling broad tasks and struggle with deterministic operations. Function-calling mechanisms like MCP may be better suited to translating 'action plans' into execution. While LLMs excel at ideation and guidance, they fall short in precise execution. For Agentic Finance to become truly reliable, we must move beyond LLMs to develop specific function-calling mechanisms, clear execution strategies, verifiability, and secure permission systems. In other words, today’s agent execution layer is underdeveloped—the AI 'brain' is smart enough, but it lacks a robust 'body.'"

Intelligent Agents as "Autopilots"
If co-pilot agents are like mentors, quantitative agents are like autopilots. They not only design strategies but also execute them—monitoring markets, testing trades, and acting at machine speed, enabling fully automated DeFi strategies.

A emerging example is SendAI. It is not a quant agent itself but a toolkit for building them. Its "Agent Kit" for Solana supports over 60 autonomous actions, including token swaps, new asset issuance, and lending management, directly interacting with major protocols like Jupiter, Metaplex, and Raydium.

In other words, it provides developers with a "rail system" to connect decision-making models to on-chain execution.

SendAI founder Yash summarized their vision:
"We believe every AI agent will have its own wallet in the future. SendAI is building the tools and economic layer for this ecosystem, enabling agents to perform any action on Solana. We’re creating a platform for context-aware agents that support long-running, persistent, and asynchronous complex tasks."

Meanwhile, other teams are making this capability more accessible. Lomen curates strategies and lets users "deploy with one click," lowering the barrier to quant automation without coding.

For advanced users who prefer custom systems, Unblinked offers an AI-powered strategy experimentation environment. It’s like Cursor for trading: users sketch strategy ideas, run and optimize them in a safe sandbox, and then decide whether to deploy real capital.

Some platforms combine multiple agents to complete tasks.

For instance, Almanak integrates "programming agents" and "backtesting agents": users describe strategies in natural language, the AI generates production-ready code, backtests it with over 10,000 Monte Carlo simulations, and delivers a battle-tested strategy.

Finally, some teams focus on real-time market advantages.

Giza’s ARMA agent dynamically allocates funds across lending protocols to maximize stablecoin yields. Instead of leaving funds in a single pool, ARMA continuously monitors rates, liquidity, and gas costs, moving assets accordingly. Its flagship agent has managed over $17 million, claiming an 83% higher yield than static holdings.

Overall, these quant agents significantly reduce time costs and expose ordinary users to strategies once reserved for professional teams. However, they also reveal the fragility of automation: agents can still "stumble" during data delays, protocol pauses, or market volatility.

In other words, they make you faster but far from "invincible."

The Challenges They Face
After spending time with current agents, you’ll notice recurring issues: they sometimes suggest actions for deprecated pools, rely on outdated data, and fail to adapt when multi-step plans go awry, repeatedly retrying the same failed action.

Permission management is also clumsy—users must either grant full wallet access or manually approve every minor step. Testing is superficial, with simulations failing to replicate real-world chaos like sudden liquidity shifts or governance parameter changes.

One of the most serious issues is that these agents largely operate as "black boxes."

Users can’t see which inputs were read, how options were weighed, whether real-time states were checked, or why a specific transaction was chosen. Without signed operation records, there’s no way to verify if "promised results" match "actual execution."

Users must "babysit" the automation—inefficient and making performance hard to evaluate.

Without mechanisms to verify decisions and prove actions adhere to stated strategies, users can’t distinguish "reliable systems" from "well-marketed hype."

For larger capital, DeFi platforms must shift from "trust us" to "verify for yourself." This is the critical turning point for building "auditable, governable, and trustworthy" Agentic Finance infrastructure.

The Infrastructure Gap
The core issue is the lack of foundational tools to ensure agents are trustworthy, consistent, and secure at scale. To fix this, we need infrastructure that verifies agent behavior, confirms execution results, and enforces uniform rules across all environments. Only then will people confidently entrust real capital.

Most users don’t care about the agent’s "thought process"—they just want confirmed outputs that are correct, verified, and within safe boundaries. For building trust, "verifiable reliability" matters more than "visibility."

This is the essence of "Verifiable Reliability." Agents don’t need to log every internal step but should operate under clear strategies and reasonable checks: spending caps, execution time windows, confirmations for critical actions, etc.

Under the hood, these rules can be enforced via trusted execution environments (TEEs) or similar systems—proving compliance without exposing all details. The result: auditable outputs when needed and operations ordinary users can trust immediately.

This verification layer need not be one-size-fits-all. Everyday use cases can adopt lightweight security and standardized metrics, while high-risk or institutional scenarios can require stronger proofs and formal verification. The key is that each infrastructure layer should provide measurable reliability matching its risk level.

Preparing Protocols for Agents
The next step is making protocols "agent-friendly."

Most DeFi protocols aren’t designed for intelligent agents. They need to provide more stable, secure execution interfaces: operation previews, safe retries, and consistent data structures for execution. Permission designs should be "scoped" rather than "all-or-nothing," letting agents operate within clear boundaries without controlling the entire wallet.

Without these foundations, even the smartest agent frameworks will be tripped up by fragile underlying layers. Once these bases are solid, users won’t need to manually monitor automation; development teams can reduce debugging time and focus on innovation; execution results across service providers will be comparable against shared benchmarks—moving beyond marketing claims.

What Needs to Change
The solution is straightforward: make agents provable and protocols agent-ready. Add a strategy layer between agents and wallets, and require all execution processes to be traceable and verifiable, not "black-box."

For example, Termina’s SVM engine is built on this idea—providing AI agents with a real Solana runtime environment for modeling, decision-making, and learning based on on-chain data. Meanwhile, protocols should open "dry-run" interfaces, clear error codes, safe retry mechanisms, consistent core data structures (positions, fees, health factors), and session-based permission controls.

When these features are implemented, users are freed from "babysitting" agents; teams reduce system failures; institutional investors finally get the safety rails and verifiable proofs they need.

Realistic Timeline
Over the next six months, "co-pilot" agents are expected to improve the fastest. Better data pipelines will enhance their reliability in everyday use.

Within a year, as testing standards strengthen, agents will coordinate execution across protocols, with humans only approving critical steps. Longer-term, as infrastructure matures, intelligent agents may blur into DeFi’s default interaction layer—no longer separate "tools" but the primary way people interact with the financial system daily.

Conclusion
"Agentic Finance" is lowering participation barriers, making automation accessible beyond experts. But to operate at scale, it needs better "foundations": real-time data, safer permission mechanisms, stronger testing systems, and more transparent execution results.

Smarter AI alone won’t solve these problems. Real progress will come from improving underlying structures.

DeFi’s next milestone is not just scaling larger—it’s about trust in automation. And that day will only arrive when AI agents evolve from "proof-of-concept demos" into truly reliable executors.