Predictive ability is crucial for human development

Prediction is a key ability in the process of human evolution. Since ancient times, humans have relied on their senses and intuition to perceive threats and opportunities in the environment, such as recognizing predator behavior patterns, finding opportunities for hunting, and anticipating seasonal food supply. These predictions are crucial for survival.

As time goes by, human predictive abilities continue to develop. We began to use tools and make plans, such as predicting the demand for planting and storing food. We also learned to interpret social cues, anticipating others' intentions and emotions. Advances in fields such as literature, science, and mathematics, along with modern statistics, computer science, machine learning, and artificial intelligence technologies, are continuously enhancing human predictive capabilities.

Prediction markets have evolved into an important economic tool that utilizes collective intelligence to forecast the outcomes of economic, political, and cultural events. Unlike traditional polls, prediction markets enhance the accuracy of forecasts through economic incentives, as participants bet real money. Certain prediction platforms attracted nearly $4 billion in bets in the 2024 U.S. election market, with their prediction of Trump's victory even surpassing that of traditional polls, highlighting the economic value of crowdsourced forecasting.

Similar developments have also occurred in the spot and perpetual contract trading sectors. From the rise of centralized exchanges to meet global cryptocurrency demand, to recent platforms offering self-custody and no KYC services while maintaining a centralized exchange-like trading experience, this field is rapidly evolving.

With the advancement of artificial intelligence and machine learning predictive models, our ability to predict events, asset prices, and volatility is significantly improving. This will propel humanity into a new stage of evolution.

The Rise of DeFi 3.0

DeFi 1.0 introduced smart contracts and decentralized applications, allowing anyone to conduct transfers, trading, staking, lending, and yield farming operations anytime and anywhere. Essentially, it involves putting crypto assets on-chain to create economic value, with representative projects including some well-known decentralized exchanges, lending platforms, and yield aggregators.

DeFi 2.0 has expanded on this basis, introducing innovative token economics and incentive distribution mechanisms aimed at aligning the interests of different stakeholders within the protocol, and has spawned emerging markets that provide alternative sources of yield.

DeFi 3.0 introduces artificial intelligence into the DeFi space. Some refer to it as DeFAI, while others call it AiFi. Its core is the integration of large language models (LLM) and/or machine learning models (ML) into DeFi products.

From simple LLM integrations ( acting as customer support/assistants, helping users utilize protocols ), to complex multi-agent/clustering and machine learning systems, AI is fundamentally enhancing DeFi products ( such as increasing trading profits, reducing impermanent loss, improving LP yields, and lowering liquidation risks in perpetual trading, etc. ).

In addition to the DeFAI abstraction layer and fully autonomous financial agents, today we will discuss how AI/ML systems and predictive models are transforming DeFi and other verticals.

Development of Predictive Systems

Neural networks and decision trees have been around since the 2000s, when some hedge funds used these systems to predict stock and commodity prices. Early stock prediction results were quite informative, with short-term prediction accuracy reaching 50%-60%, but their application was limited due to overfitting and limited data.

The subsequent rise of deep learning and big data has enabled models to handle larger datasets, such as time series data, news, and unstructured data from social media, leading to more accurate predictions and broader applications.

In the past five years, there have been breakthrough developments, with Transformer models and multimodal AI integrating more diverse datasets, such as social media sentiment, blockchain transactions, oracles, real-time news, and crowdsourced predictions from more sources. This has enabled some AI models to achieve an accuracy rate of 80%-90% in predicting event outcomes and asset prices.

As these models continue to improve, the demand for integrating predictive capabilities into DeFi systems has significantly increased. We are currently in the early stages of DeFi 3.0, witnessing in real-time some market participants combining AI/ML systems with Web3 application scenarios.

The Integration of DeFi and AI/ML Systems

A decentralized predictive model network has integrated with multiple DeFi protocols and AI agent teams, granting it predictive capabilities (, primarily focusing on cryptocurrency price predictions, such as BTC, ETH, SOL ). Its short-term cryptocurrency price prediction accuracy is reportedly around 80%.

Some main applications include:

• AI-driven vault based on USDC, utilizing predictive technology to maximize SOL trading returns. Since April 23, its cumulative return rate has been 2.4%, with an annualized yield of approximately 10%.

• AI LP Vault, utilizing predictive price data to better position liquidity ahead of price fluctuations, thereby avoiding impermanent loss.

• Collaborate with multiple teams to support trading strategies and execution for AI agents.

The incentive distribution mechanism of a certain network can help startups offset development costs, and the team uses this network to initiate product research and development, outsourcing a large amount of development work to miners. The higher the incentives, the better the quality of the miners.

Given that machine learning models and prediction systems are among the easiest tasks to quantify (, building models that can accurately predict certain things ) is one of the verticals that subnets focus on the most.

Some sub-networks focused on prediction are developing interesting applications:

• The upcoming DeFi vault will automatically allocate user deposits to high-confidence events/markets for betting. The annualized yield from early tests is reportedly over four digits.

• Continuous improvement in football/soccer predictions. Recent performance in the World Club Cup shows that aggressive betting sizes have brought a 232% return on investment. The team is also working on developing a DeFi treasury product that will adopt a more risk-adjusted approach.

• A subnet built around a highly versatile volatility prediction model. It can be used to cover various probabilities of price occurrences ( rather than just predicting future prices ), such as predicting liquidation probabilities, the survival time/liquidation time of perpetual positions, setting liquidity provision ranges and predicting impermanent loss, predicting option strike prices and expiration times within a window, etc.

These subnetworks are so appealing because they offer incentives of 2 million to over 10 million dollars in tokens each year, attracting miners to continuously improve their predictive models.

The goal is to use these incentives as capital expenditure to guide product development and achieve commercialization/productization as soon as possible, thereby earning actual returns and offsetting the selling pressure of the tokens. Some of these subnets have begun to move towards the commercialization stage.

Future Outlook

The pursuit of higher returns and lower risks will continue, prompting builders to introduce more physical assets onto the chain. Existing DeFi yield sources will continue to be optimized and will become increasingly accessible.

The prediction market will become a major source of information, with AI acting as a market maker, while experienced participants further stimulate collective intelligence. Tools are becoming increasingly intelligent, models are becoming more precise, and we have already seen some results.

The more these systems learn, the greater their value becomes. Moreover, the stronger their interoperability with other parts of Web3, the more unstoppable the overall trend.

Ultimately, everything in the crypto space is a bet on the future. Therefore, infrastructure and applications/agencies that can foresee the future even slightly more clearly—whether through collective wisdom, higher quality data, or more accurate models—will have a significant advantage.