This episode delves into the practical convergence of DeFi and AI, exploring how AI agents and sophisticated forecasting models are being built to enhance on-chain trading, risk management, and user experience, spearheaded by platforms like Mode Network and Synth.
James Ross's Background and Entry into Crypto/AI
- James Ross, founder of Mode Network and Synth, brings seven years of experience building and investing in the crypto space, primarily focusing on DeFi applications.
- His journey started even earlier with an AI-focused fintech startup in 2015 centered on data analytics and smart suggestions, giving him early exposure to AI's potential before diving deep into the Ethereum and DeFi ecosystems.
- This background provides James with a unique perspective, blending insights from both the AI and crypto domains, informing his current work building at their intersection.
The Intersection of DeFi and AI: Rationale and Approach
- James views AI not just as a trend but as a fundamentally world changing technology poised to impact every industry, including finance and crypto.
- He sees AI integrating into DeFi across multiple layers:
- User/Agent Layer: LLMs for research, strategy suggestions, and autonomous agents executing tasks.
- Data Layer: AI-powered models providing insights and forecasts for better decision-making.
- Base Layer: Exploring decentralized AI networks.
- Mode and Synth primarily focus on the application and data layers, aiming to build tools that empower users with better on-chain decision-making capabilities. James emphasizes, “no matter what industry you're in you're thinking about how AI is going to change that industry for the better.”
Mode Network: Vision and Thesis (L2 for AI Agents)
- Mode Network was founded roughly 18 months ago as a Layer 2 (L2) blockchain. An L2 is a secondary protocol built on top of a base blockchain (like Ethereum) to improve scalability and reduce transaction costs.
- The core thesis shifted towards AI agents becoming the primary way users interact with DeFi protocols. James envisions a future where users direct agents via chat or simple interfaces, rather than manually interacting with complex DeFi front-ends like Uniswap or Aave.
- Mode aims to provide the foundational infrastructure and ecosystem to enable developers to build and deploy these AI agents effectively on-chain.
AI Agents in DeFi: Current State and Future Vision
- James acknowledges the initial hype around AI agents might have outpaced current capabilities, noting some early examples were simplistic.
- He describes the evolution of agents:
- Stage 1: Basic bots, often just LLMs linked to wallets/social media, lacking deep on-chain integration (“debatably like just not AI”).
- Stage 2 (Current Focus): Co-pilots, like Mode's AI Terminal or Wayfinder, assisting users by building or executing strategies based on user commands. Users remain in control, confirming actions.
- Stage 3 (Next Wave): Autonomous agents making suggestions and executing trades/strategies within defined parameters, requiring significant transparency and user trust.
- The immediate goal is practical utility: “How can agents help people make more money?” This involves finding opportunities (signals agents like AI XBT), executing complex strategies, and managing risk.
Addressing Skepticism: Practicality of Current AI Agents
- The host raises a valid point: for experienced users, interacting with a chat agent can sometimes feel slower or add friction compared to direct manual execution.
- James counters that current agents excel at tasks users could do but likely wouldn't due to tedium or complexity. He cites Giza's agent optimizing yield across lending protocols through thousands of transactions – something impractical for most individuals.
- The value proposition lies in automation and optimization for tasks like yield farming, basis trading, or complex strategy deployment, making sophisticated actions accessible via simpler interfaces (e.g., “one-click strategy”).
The Role of Data: Introducing Synth and Forecasting
- A key challenge identified while building agents was the lack of reliable future-predicting data sources within DeFi. James notes, “throughout DeFi, you need to ask questions about the future and currently there aren't many data sources for that.”
- This led to the creation of Synth, a data layer designed to provide these crucial forecasts. Synth operates as a subnet on BitTensor.
- BitTensor is a decentralized network that incentivizes machine learning models to collaborate and compete, creating specialized AI services. A subnet within BitTensor is a dedicated network focused on a specific task, like Synth's focus on financial forecasting.
- Synth hosts a competition where over 200 ML models forecast Bitcoin's future price volatility using time series models or Monte Carlo simulations (computational methods using repeated random sampling to model probability distributions of outcomes). Accuracy determines rewards.
BitTensor Deep Dive: Subnets, Incentives, and Mechanics (Synth's Foundation)
- Synth leverages BitTensor's incentive mechanism. Data scientists ("miners") submit forecasts to the Synth subnet.
- Their predictions are continuously scored for accuracy against real price movements. Top performers climb a leaderboard and earn rewards.
- Rewards are distributed in TAO, BitTensor's native token. James estimates the annual reward pool for Synth data scientists is potentially around $1M.
- BitTensor recently moved to Dynamic TAO (DTA), where each subnet potentially gets its own token (like an "Alpha Token" for Synth), which miners earn and can swap for TAO. TAO emissions still flow to valuable subnets based on market dynamics (relative value/market cap of their specific token). This complex system aims to align incentives and reward useful AI work.
- For Mode/Synth, BitTensor provides the framework and incentive layer, attracting data science talent without requiring Mode to directly fund all the rewards initially. The value of the subnet (and its TAO rewards) depends on the usefulness of its output (the forecasts).
Comparison with Numerai
- James acknowledges similarities to Numerai, another platform running data science competitions for financial predictions. He participated in Numerai back in 2017.
- Key differences highlighted:
- Numerai focuses on the stock market using encrypted data, with outputs feeding into a somewhat opaque hedge fund.
- Synth focuses on crypto (starting with BTC volatility), uses an open competition model inspired by Jane Street, and provides its output directly via an API for agents and users.
Synth's Application: Mode AI Terminal and Mode Trade
- Synth's forecasting data is integrated directly into Mode's ecosystem.
- Users can query Synth via the Mode AI Terminal (a chat interface) to get volatility forecasts, liquidation probability estimates, and risk assessments for potential trades.
- This information helps users make more informed decisions, for example, choosing appropriate leverage based on expected volatility (“The mode AI terminal will say something like you should go like 2x long because we're coming into a volatile period.”).
- Trades suggested or configured in the AI Terminal can then be executed on Mode Trade, a perpetuals DEX built on Mode using Orderly Network as the backend engine and liquidity source. Orderly Network provides shared liquidity across multiple chains it's deployed on.
Broader DeFi Implications: Risk Management, Options, and Capital Efficiency
- The host and James discuss the potential for Synth's volatility data beyond simple trading, particularly for risk management and options markets.
- Volatility is a key input for options pricing. James confirms they are using Synth data internally to price options and explore integrations with options protocols.
- They acknowledge that options remain underdeveloped in DeFi compared to perpetuals (perps), possibly due to complexity, but represent a significant opportunity for sophisticated risk management and capital efficiency, similar to their role in traditional finance.
- Bringing better risk management tools (like volatility forecasts) to retail traders is seen as crucial for user retention on trading platforms.
Future Directions: Real-World Assets and Expanding Synth
- A key next step for Synth is expanding its forecasting capabilities beyond Bitcoin to include equities and indices.
- This aligns with a broader trend of bringing Real-World Assets (RWAs) – tokenized versions of traditional assets like stocks or bonds – onto blockchains.
- Mode aims to list these new assets on Mode Trade (as perps) and potentially across other DeFi applications within its ecosystem, increasing access for global users. James sees this as fulfilling DeFi's promise of open access to financial opportunities.
Competitive Landscape and Mode's Positioning
- Mode Network operates as an L2 increasingly building its own native applications (like Mode Trade), similar to models seen with Hyperliquid (L1) or Nosis Chain.
- While supporting ecosystem builders (like Giza, Developer DAO), Mode is carving a niche focused on AI-driven trading tools and data integration via Synth.
- James mentions other players in the agent space like Wayfinder and Hey Anon, acknowledging they are taking different approaches, while Mode focuses more on advanced strategies and forecasting data.
User Adoption and Current Use Cases
- Adoption is still in its early stages, but James reports seeing users actively engaging with the AI Terminal, querying forecasts, and using the tools.
- The target user persona is currently DeFi-savvy individuals interested in new tools and protocols.
- The long-term goal is to leverage AI tools to educate users, improve their trading performance, manage risk better, and ultimately increase retention – a critical factor given high user acquisition costs in crypto.
Impressive and Less Impressive AI Projects
- James expresses skepticism towards early "AI" projects that were merely LLMs connected to Twitter accounts with wallets, calling them potentially a “meme of AI” and lacking real on-chain substance.
- He highlights impressive work within the BitTensor ecosystem, mentioning Codex (building tools and on-chain trading agents like Billy Agent) and Score (using sports forecasting subnet data for an agent trading on sports markets).
Market Outlook and Final Thoughts
- Acknowledging the current market downturn, James remains optimistic about the builders in the space. He states, “market caps of everything is down like significantly. I think we'll see the people that continue building rise up again.”
- He encourages listeners interested in Mode or Synth to follow their progress via Twitter (@JRossTreacher / @ModeNetwork).
This discussion highlights AI agents and predictive data (like Synth on BitTensor) moving beyond hype to offer tangible tools for DeFi trading and risk management on platforms like Mode. Investors and researchers should actively monitor agent capability advancements and the integration of RWAs for emerging strategic opportunities.
