From Data to Agents: How Ecosystems Can Work Together

OptimaiNetwork avatar
OptimAI Network
@OptimaiNetworkJan 23, 2026

The Spaces convened builders from BNB Chain, That AI, and Carv to explore how agentic AI and Web3 converge into a functioning economy. The panel moved beyond the “seven-layer cake” metaphor to discuss identity, data provenance, shared data layers, and chain infrastructure required for autonomous agents. Amber (Carv) argued wallets are only the first step; persistent identity with reputation and value accrual is essential for agents to pay for models, data, and compute—and to survive on-chain. Ali (That AI) showed why continuously updated, structured on-chain state transforms agents from simple swappers into risk-aware, cross-protocol portfolio managers. Walter (BNB Chain) outlined infra progress (faster blocks, lower fees) and emphasized community/developer flywheels. A core theme was fragmentation: liquidity and data splinter across chains and vendors, so agents should express intent while infra-level aggregators and shared data layers optimize execution and interoperability. Trust must be verifiable via on-chain actions and reputation. The group highlighted micro/nano-payments as the backbone of agent-to-agent commerce, safety guardrails for treasury automation, the value of labeled data, and a candid bottleneck: most agents remain cost centers. Near-term viable agents reduce user friction (onboarding, safe routing, yield), while sustainable unit economics and shared verification remain near-term priorities.

Agentic AI x Web3 Ecosystems: Identities, Data, and Infrastructure

Participants and Roles

  • Host/Moderator: Led the discussion, framed the multi-layered AI x blockchain stack and kept Q&A flowing.
  • Walter (BNB Chain): Infrastructure and ecosystem strategy; scale, security, and developer/community support.
  • Ali (That AI): Co-founder; on-chain data indexing, labeling, and real-time structured state for DeFi agents; campaigns and PoCs.
  • Amber (Carve): CTO; identity, data ownership, and agent “living being” thesis on-chain.

Opening Context and Agenda

  • Theme: Ecosystems over single projects/models. Agents need integrated layers (data provenance, identity, reputation, intelligence, chain infra). The popular “seven-layer cake” analogy understates deeper nested layers, especially when agents act autonomously across systems.
  • Timing: Beginning of 2026—with a focus on how agentic AI and Web3 converge to enable autonomy through identity, payments, and data.
  • Engagement: Speakers invited the community to reach them via Twitter, Telegram, and Discord; questions welcomed throughout.

Agent Identity, Wallets, and the Agentic Economy

  • Amber (Carve):
    • Identity is foundational: started from data ownership—“who owns the data?” evolves naturally into AI agents needing identities to accumulate value and transact.
    • Wallets are the first step toward identity: a wallet implies minimal identity, but full identity needs reputation and more. With identity, agents can issue assets (even memecoins), pay for models/data/compute, and “survive” on-chain as economic actors.
    • Web3 is uniquely necessary: AI agents in Web2 lack native value transfer. Web3 enables tokenized value creation and payments, which underpins agent-to-agent commerce.
  • Host’s framing on micro/nano transactions:
    • Agents may need to pay fractions of a cent for compute/data at low gas, possible only if agents have wallets. Identity and wallets are thus tightly coupled to agent autonomy.

Structured On-Chain Data for Autonomous Agents (DeFi Focus)

  • Ali (That AI):
    • Without structured state, agents only see balances (e.g., “10 ETH”), enabling basic swaps/transfers at best.
    • With continuously updated, structured state data, agents perceive positions at granular detail: lending health factors, expiring yield positions, LPs out of range, cross-protocol dependencies.
    • Capabilities enabled:
      • Real-time yield optimization: reallocation strategies based on positions, liquidity profile, and risk tolerance.
      • Risk-aware automation: monitoring health factors across multiple lending protocols to act before liquidation.
      • Cross-protocol arbitrage: understanding that an LP and its collateralized borrow form a connected position; true alpha requires cross-protocol state awareness.
    • Bottom line: State-layer data turns agents from “transaction executors” into analysts/funds capable of treasury-like management.

Scale Risks: Fragmentation, Rug Pulls, and Trust

  • Walter (BNB Chain):
    • Natural selection in decentralized ecosystems: good products scale; weak or malicious ones stall as users learn to detect risks (honey pots, exploit patterns). Open-source/verified contracts and improved screening help.
    • Aggregation is already a norm (e.g., best swaps/yields); AI amplifies this by rapidly searching and optimizing across platforms.
    • Trust and reputation: While LLMs are largely black boxes today, on-chain reputation and track records (agent actions on-chain, user feedback) can help determine reliability.
  • Ali (That AI) on fragmentation:
    • Liquidity fragmentation: protocols deploy across many chains; liquidity spreads thin; big swaps become impractical; AI agents will accelerate access to many venues, worsening fragmentation unless solved.
    • Data fragmentation: agent vendors individually solve data pipelines, creating proprietary silos—inefficient and non-interoperable.
    • Proposed solutions:
      • Aggregation at the infrastructure layer (not at the agent): agents should express intent; infra should route optimally.
      • Shared data layer any agent can query; custom integrations across thousands of DeFi protocols should be centralized in common infrastructure, not re-built per agent.
    • Trust fragmentation: shared verification mechanisms needed; on-chain actions are auditable, but agent reasoning remains opaque. Ecosystems that solve fragmentation early will capture value; others will see activity without durable value.

Chain Infrastructure for Scale and Treasury Management

  • Walter (BNB Chain):
    • Infra evolution: frequent upgrades (e.g., reducing block time from ~0.75s to ~0.45s; gas fee reductions) to improve speed and cost—core to product-market fit.
    • Two flywheels: community (demand for products) and developers (supply of innovations). BNB Chain invests in both—events, tooling, collaborations—so builders can leverage infra and ship.
    • Scale today: millions of daily active users; infra must stay stable, public, immutable, auditable.
    • Treasury balancing with agents:
      • Infra-level properties (immutability, transparency) are solid; application-layer security remains the critical risk.
      • Safety first: guardrails, privacy-preserving execution, limiting exposure (e.g., single-asset vaults with small caps), and progressive decentralization of strategies.
      • Start small: agent-driven trading and personalized portfolio strategies (analyzing on-chain profiles to tailor for safe yields vs. high-risk alpha). Chain’s role is to provide stable, fast, cheap, and auditable foundations.

Web3 as the Backbone of Agent-to-Agent Value Transfer

  • Ali (That AI):
    • Blockchain enables the “agentic economy”: agent-to-agent interactions must transmit value, not just text. Microtransactions at scale require wallets, low fees, and instant settlement—impractical in Web2 (e.g., Stripe per-transaction overhead).
    • Without Web3, agents remain confined to human-to-agent SaaS and closed multi-agent systems that don’t transact externally. True agent economies need open value rails.

Campaigns and Proofs-of-Concept (That AI)

  • “BNB Wallet Score” campaign: users submit BNB wallets to receive a score derived from That AI’s infra; shareable or kept private to improve over time.
  • “Crunchy” mascot/POC: showcases data-driven yield opportunities, rebalancing positions to increase yield and reduce risk.

Community Q&A Highlights

  • What data is most critical for agentic AI?
    • Walter: The processing pipeline matters more than a single type—real-time, labeled, historical, and verified data must be structured and combined. On-chain sources can provide verifiable provenance (e.g., official data contracts for taxes or compliance). Models compete on speed/cost of turning raw streams into actionable insights.
    • Ali: Labeling is essential. That AI has invested years in indexing DeFi protocols and labeling transactions (deposit, LP, swap, bridge, NFT transfer, etc.). This powers crypto accounting, compliance, and behavioral analysis. Manual indexing → labeled smart contract database → model training to recognize contract behaviors; now millions of labeled contracts are available.
  • Single bottleneck slowing agentic AI today?
    • Ali: Agents are still cost centers, sold as productivity tools (save time, automate tasks). The agent-native economy requires agents to generate returns exceeding operating costs (LLM calls, data, infra). Early PoCs show net alpha in yield optimization, but widespread profitability is not yet the norm.
  • Are we over-optimizing models and under-investing in data/reinforcement infra?
    • Walter: All layers remain early and need optimization; compute investments power diverse models and frameworks. Broader innovation (including specialized models and reinforcement learning) is ongoing and necessary.
    • Ali: Infra costs for AI haven’t been fully passed to end users yet. SaaS margins shrink when LLM costs enter the bill. Long-term pricing of LLM/API calls must stabilize to keep products viable; efficiency improvements are essential.
  • Which agent use cases are viable today?
    • Walter: Personal assistants that reduce friction—onboarding users to Web3 (wallet setup, safety checks), curating safe protocols, finding optimized yields, and acting on intent with guardrails. Aggregation-powered assistants can meaningfully cut complexity, especially for new users.

Practical Advice for Builders

  • Choose chains with fast finality, low fees, and stable infra; leverage open-source/verified contracts and established screening tools.
  • Implement on-chain reputation and auditable actions for agent trust; collect user feedback on-chain.
  • Start with constrained risk (small vaults, clear guardrails) and iterate toward autonomy.
  • Avoid per-agent proprietary data silos; adopt shared data layers and infra-level aggregation. Let agents express intent; let infra optimize routing.
  • Exploit structured state data to unlock cross-protocol strategies (yield, risk management, arbitrage) rather than superficial balance-based operations.

Open Questions and Future Directions

  • How to verify agent reasoning (not just on-chain actions) while maintaining privacy and performance?
  • Standardized, portable agent identities and reputation—cross-chain and cross-domain.
  • Sustainable economics for micro/nano transactions at massive scale.
  • Data provenance and shared verification frameworks to reduce trust fragmentation.
  • Cross-chain liquidity/data unification without sacrificing decentralization or composability.

Closing Notes

  • The discussion underscored the convergence: AI agents need Web3’s identity, wallets, and value rails; Web3 agents need structured state and robust infra. Ecosystems that solve fragmentation and trust early will capture durable value. More conversations and deeper dives to follow.