Decoding 2026: The Explosion of AI Agents & The Next Alpha in Web3

MetaEraHK avatar
ME
@MetaEraHKDec 30, 2025

The Spaces explored how AI agents are reshaping Web3, anchored by Holmes AI (building sovereign, on-chain digital personas) and Deal (a modular L1 focused on scalable, verifiable data). Carrie moderated, with Kai (Holmes AI co-founder/CTO) and Ted (Deal co-founder) sharing 2026 forecasts: prediction market users could grow 20x, agents 1000x, and on-chain data markets emerge; agents may generate more on-chain transactions than humans. The core thesis: decentralized networks solve AI’s trust and data auditability through verifiability and immutability. Holmes AI chose Deal for low-cost, high-throughput storage and per-app execution scalability to log agent behavior transparently. Both see intertwined opportunities across infra and applications, with agent-driven automation and data monetization likely leading, followed by stronger, cheaper infra. Milestones include Holmes AI’s possible Q1 2026 TGE, continued airdrops with fairness mechanisms, and Deal’s mainnet and bridges. Tokenomics reward high-quality data providers, agent builders, and governance roles (data quality checkers), while Deal offers validators, mini-pools, and ecosystem pools. Q&A covered scaling to 10M users (Deal confident), practical security via decentralized data layers, and the primacy of purposeful, human-centered agent use. The session ended with giveaway winners and community calls to follow both projects.

AMA Summary: AI Agents, Web3, Holmes AI x Deal Network (Twitter Spaces)

Session Overview

  • Theme: The rise of AI agents and how they reshape Web3, moving toward 2026; focus on decentralized AI infrastructure, agent-driven economies, and data markets.
  • Host/Moderator: Carrie.
  • Speakers:
    • Kai (KY): Co-founder and CTO of Holmes AI (sometimes referred to as Home CI/Homos AI in the session).
    • Ted: Co-founder of Deal Network (a modular, shared-architecture L1).
  • Audience: Builders, investors, and users interested in AI x Web3.
  • Incentives: Live Q&A winners receive 10 USDT; announcement of five winners from last week’s $75 campaign.

What Holmes AI Is (as introduced by the host)

  • Positioning: A “digital Avatar layer” of AI—an on-chain, intelligent digital person (agent persona) that users fully own, control, and monetize.
  • Backing & Traction: Backed by top-tier investors (as referenced: HashKey, Bitrise, Waterdrip). Raised $8M to date; pre-TGE; 1.5M+ users already onboarded at the time of intro (later updated to ~2M users).
  • Differentiation: Emphasis on full sovereignty over AI personas—data ownership and value capture—enabling an agent-driven economy in Web3.

Opening Predictions for 2026

  • Kai (Holmes AI):
    • Prediction market users to grow 20x.
    • Number of AI agents operating in prediction markets to grow 1000x.
    • An on-chain data trading market of comparable scale will emerge.
  • Ted (Deal):
    • By end of 2026, AI agents will generate more on-chain transactions than humans.

Why Top VCs Are Increasing Exposure to AI Agents (late 2025)

  • Kai’s view:
    • Despite a “cold” crypto market, several Web3 vectors are accelerating:
      • Rising volumes in prediction markets.
      • More autonomous agents launching.
      • Rapid growth of stablecoin payments for online services.
      • Users buying on-chain “stocks”/assets.
    • 2026 will present huge opportunities in agents and data markets; the sector remains early and both Web2 and Web3 are still digesting AI’s power to find best-fit verticals.
    • Stablecoin adoption lowers barriers, expanding the crypto user base.
    • Data monetization will become mainstream: individuals (e.g., domain experts like sports analysts) will sell useful data to agents/users for decisions (e.g., prediction markets).
    • Trend toward user-owned media/data monetization reinforces the opportunity.
    • Net: Larger user base + valuable data + agent proliferation → massive opportunities in agent economies and data trading in 2026.

Trust, Data, and Why Decentralization Matters (vs Web2 AI)

  • Ted’s argument:
    • Verifiability: In Web2, you cannot audit how AI agents reach conclusions (e.g., a trading decision). Decentralized networks can post inference audit trails or decision trees on-chain for third-party verification.
    • Immutability: Centralized providers can alter/erase data; decentralized networks ensure data cannot be retroactively modified and remains retrievable.
    • Deal’s role: Provides data and consensus layers (and execution) with high scalability and throughput; can support extremely large node sets while maintaining performance.
    • Outcome: Decentralized infra ensures transparent, tamper-proof, and accessible data/inference pipelines—critical for agent trust in financial or high-stakes use cases.

2026 Opportunity Landscape: Infra vs Applications

  • Kai (sequence of emergence):
    • Agent-driven automation first.
    • Then advanced data layers that make data tradable.
    • Then larger/cheaper infra as user scale demands it.
    • 2026 focus: Prediction markets and on-chain data trading due to:
      • Stablecoin payments lowering onboarding friction.
      • Proven traction of platforms like Polymarket, Cozi (as cited).
      • Compliance tailwinds (Kai referenced a potential IPO event around a prediction market in the US; presented as his cited view).
      • Prediction markets are complex: success requires strategy + massive information processing. As AI becomes ubiquitous, valuable data becomes the true moat. On-chain is well-timed for information monetization.
  • Ted:
    • Both infra and apps will be big; they co-depend.
    • Current infra costs are too high for agents’ micro-payments/micro-work at scale; significant room to cut costs and increase performance.
    • Today’s agents mostly function as co-pilots embedded in legacy workflows. Future: agent-centric workflows/products across work, personal life, and entertainment—bigger growth than current chatbot-like patterns.
    • If forced to choose: Infra will lead (as enabler and value capture via gas and network effects).

Why Holmes AI Built on Deal Network

  • Kai’s rationale:
    • Evaluated L1s across performance, cost (especially storage), security, decentralization, and ecosystem.
    • Deal stood out with highest performance and lowest storage cost; EVM compatibility.
    • Holmes AI needs to store more than typical trading data: agent activity logs, provenance/audit trails—data must be transparent and queryable to assess agent behavior and output quality.
    • Traditional public chains (e.g., Ethereum, Solana—as referenced) are optimized for transactional data; Holmes AI needs a data-first chain for large-scale, low-cost write/read.
  • Ted’s complement:
    • Scalability: Modular/shared architecture with ability to allocate dedicated execution spaces for apps; “infinite scalability” by spinning additional execution environments as needed.
    • Throughput: Referenced support “up to 100k TPS” and “~10 MB/s per execution shard.”
    • Data architecture: Sharded data sub-networks ensure performance without sacrificing availability; 24/7 reliable, low-cost storage.
    • Path to scale: From thousands to millions of agents without performance degradation.

Roadmap and Milestones (Q1 2026 and Beyond)

  • Holmes AI (Kai):
    • TGE targeted for Q1 2026 (January/February window mentioned).
    • Airdrop program continues until TGE; more giveaways imminent.
    • User growth: ~2M users after foundation announcement; implementing fairness strategies to reward early believers, real humans, and high-quality participants.
    • Product: Collaborating with prediction market agent builders; aim to get agents online early.
    • Pre-TGE: Applying to list with leading centralized exchanges (first preference mentioned as “Binance (Alpha)” and possibly another CEX; total listings limited to no more than two at launch, per Kai’s remarks).
    • Additional plans: NFT sales; staking of stablecoins on-platform may increase airdrop allocations; NFTs may be required for governance roles (data quality checker, agent builder) and for trading personas.
  • Deal Network (Ted):
    • Near-term (2026): Mainnet launch; trust-minimized bridges to other ecosystems; onboarding more validators.
    • Mid-term (1–3 years): Grow validator set and develop the ecosystem to create strong network effects and incentives for DApps to deploy on Deal.

Tokenomics and Incentives

  • Holmes AI (data-centric incentives):
    • Core thesis: Better agent outputs require high-quality, well-structured personal/contextual data; Holmes AI processes user data to build personas that agents leverage to deliver superior outputs.
    • Rewards go to:
      • Data providers: Largest share of token distribution reserved for high-quality personas and data; rewards scale with data quality and usage (popularity) of personas.
      • Agent builders/partners: Especially those building prediction market agents and other high-utility agents.
      • Governance roles: Data quality checkers and other governance participants receive ongoing rewards for work that assures data integrity and model reliability.
    • Sustainability: Ongoing airdrops post-TGE tied to data value/usage; governance tasks rewarded per task (analogy: akin to mining rewards, but for data validation and governance labor).
  • Deal Network (network participation):
    • Validators: Support both full and light validator modes; low entry barrier for light validators; full validators for higher commitment.
    • Mini-pool staking: Stake tokens without running infra; pool manages staking and distributes rewards.
    • Ecosystem launch pool: Deal’s ecosystem/foundation allocates tokens to apps launching on Deal; DApps may share tokens with Deal stakers, benefiting both network and app communities.

Community Q&A Highlights

  • Challenges for AI agents in Web3 (scalability, security, adoption) and Holmes AI’s approach:
    • Kai: Holmes AI functions as a data layer built on Deal; relies on Deal’s on-chain security and data/transaction integrity. Architecture is dApp-like and inherits L1 security guarantees.
  • Vision: Personal AI clones interacting as a network of digital people?
    • Kai: Skeptical of AI-to-AI chatting without purpose—unlikely to be valuable. AI should remain a tool to augment humans; human-to-human interaction remains primary. Agents must deliver concrete utility (search, decision support, output generation) rather than self-referential interaction.
  • TGE timing, CEX listings, marketing, sustainability:
    • Kai: TGE Q1 2026 target; aiming for top-tier CEX listing(s), potentially “Binance (Alpha)” plus one more. Post-TGE sustainability driven by continued data-provider rewards, governance incentives, and connecting valuable data to high-performing agents (e.g., prediction market agents).
  • Post-TGE plans for community:
    • Kai: Token emissions continue to reward high-quality data contribution and governance work (data quality checks). Rewards scale with data usage/popularity—designed to keep contributors engaged long-term.
  • Tokenomics and long-term holder value with tradable personas:
    • Kai: Value accrues from Holmes AI’s ability to transform user data into model-ready formats, improving agent outputs and delivering utility (trading, prediction, search). Connecting valuable data with effective agents creates demand and monetization pathways.
  • Can Deal support Holmes AI at 10M users?
    • Ted: Yes, expected to be fine. Traffic patterns for such apps tend to be smooth. Deal can scale horizontally by adding execution environments; per-shard throughput ~10 MB/s, plus dedicated application execution layers.
  • Data sources for Holmes AI agents and data quality enforcement:
    • Kai: Agents learn from user-contributed context—chat logs (e.g., WeChat), purchase history, browsing/news history, sites visited. Platform auto-injects relevant persona context into model prompts to improve output quality. Data quality is incentivized via tokenomics and governance roles (data quality checkers) to ensure integrity and utility.

Announcements and Rewards

  • Live Q&A: Selected audience questions awarded 10 USDT each; details to be posted on official Twitter within ~10 minutes of session end; winners to DM to claim.
  • Past campaign winners (last week’s $75): Announced five winners (as read by the host):
    • Alex Burkett
    • Siki Melt a14
    • chungs cp
    • Zhao 280294
    • Liu hao 096

Notable Numbers and Technical Claims (as stated by speakers)

  • Holmes AI: Raised $8M; pre-TGE; onboarded from 1.5M to ~2M users.
  • Deal Network: Claimed support up to ~100k TPS; ~10 MB/s per execution shard; scalable data and consensus layers; sharded data sub-networks; can host very large validator/node sets.

Key Takeaways

  • 2026 is set up as a breakout year for agent-driven applications and on-chain data markets, with prediction markets cited as a major near-term vertical.
  • Stablecoin adoption and data monetization are core drivers of user growth and on-chain activity.
  • Trust in AI agents hinges on verifiable, immutable inference and data trails—decentralized infra is positioned as the solution.
  • Holmes AI focuses on persona quality and data sovereignty, rewarding users and builders who contribute high-value data and agents.
  • Deal Network positions as a high-throughput, data-first L1 with modular scalability and low-cost storage, aiming to underpin large-scale agent economies.
  • Both infra and apps will matter; infra cost/performance improvements are essential to unlock micro-transactions and agent-centric workflows at scale.