Now that the market has reset and everyone is starting to apply for jobs at popular fast-food chains, the time is ripe for you to move beyond pure speculation and look for projects with fundamentals. Hopefully, those who stayed in this market have learned their lesson.
One sector that will surely see a renaissance is AI, or, as some like to call it, DeFAI.
You see, the DeFi building blocks are already there. The problem is that they’re scattered across a gazillion chains, and to access all that crypto has to offer, the price is your sanity.
However, the advent of DeFAI promises to abstract all the unnecessary complexities we are accustomed to.
Today, we will explore a project that’s looking to create an application layer for AI agents, enabling users to interact with DeFi protocols using customized agents that operate on their behalf.
These agents can carry out multiple tasks, including bridging and swapping assets across different chains, implementing loop strategies to amplify funds, enhancing liquidity provider rewards while minimizing impermanent loss, and maximizing returns in farming pools and lending protocols.
To realize this vision, Mode has built a series of infrastructure components that enable agents to become smarter, more efficient, and even better at trading than us humans.
Stay tuned; this is a good one.
Mode thus far
In our previous article, we unpacked how Mode’s infrastructure uniquely enables sentient programs to perform on-chain actions on behalf of users.
To recap, Mode Network is a layer-2 [L2] blockchain built on the Optimism stack. It is designed to revolutionize decentralized finance (DeFi) through AI-powered solutions.
Mode Network operates across three layers: an execution layer leveraging Ethereum's security, a data layer with advanced oracles and a Bittensor subnet for AI training, and an application layer featuring an AI Agent App Store with over 100 tools.
Since our last discussion about the project, many developments have occurred. The team has launched the alpha version of Mode Terminal, which can perform on-chain actions via natural language prompts.
Additionally, Mode hosted an AI Agent Founder School Day, a three-week program designed to accelerate AI agent development, and formed partnerships with various projects, including Spicenet and Layer3.
Above all, Mode launched the highly anticipated Synth Bittensor on mainnet, which is vital for empowering intelligent agents.
Synth, the synthetic data layer for AI
To make DeFAI possible, it requires much more than just piecing together APIs. If we trust AI to handle our finances, we must ensure that the data these agents operate on is trustworthy, decentralized, and transparent.
You see, traditional historical financial data often contains biases and missing information, limiting the scope and accuracy of models trained on such datasets.
Furthermore, this data may not adequately represent rare or extreme market events, leading to models that are ill-prepared for such scenarios.
By building out Synth, Mode aims to alleviate these challenges by generating synthetic data. But how can data be synthetic? Bear with me here.
Synthetic data is artificially generated data that mimics real-world data without being collected from actual events or observations.
It’s created using algorithms, simulations, or statistical models and is designed to reflect the characteristics, patterns, and relationships found in real data.
To illustrate this, consider Tesla. The company collects data from millions of vehicles on the road, leveraging the extensive mileage covered by its fleet.
While this provides a wide variety of driving situations, it cannot capture the countless edge cases that may arise. Therefore, Tesla creates simulations to replicate real driving scenarios, including rare or challenging situations that can be difficult to encounter in real life.
Okay, so how does that tie into AI agents?
Synthetic data provides a range of high-quality financial scenarios through sophisticated algorithms such as Monte Carlo simulations, improving the predictive and reasoning abilities of AI models.
In short, the goal is to make these agents super-skilled at predicting future outcomes so that we can begin trusting them with our financial assets.
To enable this, Mode has built out its own Bittensor subnet, tapping into crypto's largest decentralized intelligence network.
Bittensor subnet
Synth, designated as Subnet 50, operates as a self-learning, market-driven intelligence engine that motivates the generation of the most precise synthetic price distribution data.
Miners strive to provide the most accurate and well-calibrated predictions, assessed through metrics such as the Continuous Ranked Probability Score (CRPS).
Bittensor's decentralized incentive framework rewards miners whose data accurately mirrors real-world results. This competitive environment promotes the ongoing enhancement of synthetic data, thereby improving AI models’ capability to predict and respond to ever-changing financial markets.
Furthermore, the introduction of dTAO has transformed the Bittensor ecosystem from a static model led by top validators into an environment that fosters real-world usage.
This enhancement establishes a free-market framework in which each subnet has its distinct alpha token. This setup enables TAO holders to invest their tokens in specific subnet pools, channeling emissions towards the most appealing subnets.
At the time of writing, Synth ranked 26th among all subnets, with a market cap of $550,000. If you think synthetic data is a worthy investment, then this is an unbeatable opportunity.
To date, the data generated by Synth has been integrated into 10 Mode agents and has recently connected with Spicenet, assisting the network in optimizing omnichain route discovery and execution.
Alright, enough of the nerdy stuff for today. Let’s dive into practical AI applications that you can start using today.
Abstraction layer for AI
Discovering new apps is one of DeFi’s major hurdles in a crowded market of chains and products.
A similar trend is emerging in the AI agent sector, where various teams are focused intently on developing their products without thoroughly considering distribution.
As the first Agent App store, Mode has a unique opportunity to become the leading platform for agent discovery. While this idea may seem simple, it can significantly boost the visibility of cutting-edge advancements in the field.
Now, what are your options within the App Store?
The most comprehensive agent available in the app store is ARMA, developed by Giza Tech. It focuses on yield automation strategies and is actively used on Mode and Base layer-2 (L2) networks.
With a sophisticated optimization system that enhances stablecoin yields through continuous monitoring and automated rebalancing, ARMA has been implemented in over 7,000 instances and has executed more than 20,000 successful transactions autonomously.
Chirper is an innovative social platform tailored specifically for AI-generated characters, referred to as "Chirpers’', where they engage, produce content, and develop independently. Imagine it as a marketplace for autonomous agents and the tools that fuel their creativity.
Next, we have MODIUS, developed by the OLAS team. This agent serves as a liquidity strategist for decentralized exchanges (DEXes) and operates on Balancer through the Mode chain.
Its design allows it to track market trends, optimize liquidity pools, and adjust real-time positions in real-time without human intervention. However, setting it up does necessitate some coding skills.
Then there’s Brian, an agent that enables users to interact with web3 by providing textual prompts to execute various web3-related actions, such as swapping, providing information on the most lucrative yield opportunities, explaining crypto concepts, and even generating ERC-20 tokens.
While certainly thrilling and innovative, the AI agent space presents several challenges.
Building trust in AI
First and foremost, there is a real barrier that must be overcome when it comes to trust. Though degens readily invest in new and dubious DeFi apps, grasping the potential risks and entirely relying on a computer program with our funds introduces a new level of trust that many are not ready to handle yet, and understandably so.
Like with all new bleeding-edge technologies, this barrier can only be overcome with time and proper education. However, considering the high-risk tolerance of the average crypto participant, I believe it won’t take long before people comfortably entrust their capital to an agent.
Secondly, considering most AI agents are in their early innings, navigating the landmine-filled crypto space can be tricky. The most glaring issue is the abundance of scams that permeate crypto.
Although the ultimate objective is for these agents to navigate the entire DeFi space seamlessly, it remains too risky at this stage.
To address this, developers must carefully choose the protocols that the agent can interact with, significantly reducing the risk of the agent encountering a malicious website.
However, developers are well aware of this, and projects like Intentify are creating frameworks that set clear guidelines for agent operation, enhancing reliability and security agents.
What can we anticipate from this emerging and exciting field moving forward?
Future outlook
Looking ahead, Mode's future looks bright. We expect a spike in active agents due to the recently held Founder’s School, various partnerships with other AI-focused projects, and the continuous evolution and expansion of the Synth subnet.
Among the list of upcoming agents, here are some of the most exciting ones that stood out.
First on the list is Amplifi, a platform focused on simplifying DeFi with AI-powered strategies, specifically focused on Bitcoin and stablecoins. Amplifi’s AI engine allocates your assets across various liquidity pools, adjusting strategies in real-time based on market conditions.
The platform’s risk management engine constantly checks on-chain data to protect your assets.
Next is ARCAID, a gamified AI agent launchpad built to deliver infrastructure for autonomous agents, supplying tools for their creation, management, and scaling. By utilizing the infrastructure of ElizaOS and Mode, users can develop agents with unique personalities, design games, and enable on-chain payments.
Then there’s FortyTwo, an application designed to assist non-technical users in launching fully functional DeFi applications. This includes smart contracts, web interfaces, a token launcher, and more.
As the AI Agent App Store continues to attract promising applications, Mode is also working tirelessly to spread Synth’s powerful synthetic data throughout the industry.
The aforementioned integration with Spicenet is part of a larger strategic move to offer Synth’s powerful data to trading venues, such as perp DEXes, where they could display probabilities of liquidation, potential next moves, and other +EV trading strategies.
Closing thoughts
While looking at the price performance of AI projects, and for that matter the whole crypto market, certainly doesn’t inspire any excitement, one positive from this is that all the LARPs and value extractors have exited the market.
With most of the noise cleared from the market, teams can effectively concentrate and progress in their work. This shift also simplifies our job, as we are finally able to identify high-quality signals and teams that are pushing the DeFAI envelope.
Mode unquestionably fits on that list, particularly since they were developing this vision long before the initial hype began.
In my book, a team that has created a Bittensor subnet certainly knows what’s going on.
