Original author: ChandlerZ, Foresight News

If human society has always harbored curiosity and a penchant for the future, then crypto-native prediction markets are transforming this age-old need into a quantifiable, liquidable, and reusable public good. Over the past decade, the democratization of information was accomplished by the internet; in the Web3 and crypto spheres, value and belief are also being tokenized and priced, forming a more verifiable and incentive-compatible form of value democratization. The integration of AI elevates prediction from simple price feeds to more complex judgments and decisions, giving prediction an infrastructure-like quality. Excluding speculative interpretations, prediction markets serve as the underlying information foundation for governance, hedging, and resource allocation. Google's decision to integrate Polymarket and Kalshi's market probabilities into Google Finance in November 2025 signifies that prediction data is entering the public network layer with access to hundreds of millions of users. This is both an endorsement of the industry and a signal of increasing demand.

Why is the prediction market a crucial battleground for Web3?

The essence of prediction markets is to aggregate tacit knowledge scattered in individuals' minds into public probabilities through prices. This idea can be traced back to Robin Hanson's Futarchy, where prediction markets are set as the primary mechanism for information aggregation in a system where value objectives are determined by voting and factual judgments are left to market pricing. Academic research also shows that prediction markets outperform simple polls in many scenarios in depicting event outcomes, especially in terms of dynamic updates and incentive constraints.

However, if we shift our perspective from theoretical reasoning back to the real market, we'll find that this mechanism of aggregating knowledge through price is being voted on by capital and users in 2024-2025. Prediction platforms like Polymarket and Kalshi have repeatedly seen daily trading volumes approaching or even exceeding $100 million, with cumulative trading volume leaping to the tens of billions of dollars, marking the shift of prediction markets from niche experiments to full-blown explosion. Data shows that Polymarket's monthly active traders reached a record high of 477,850 in October, surpassing the previous record of 462,600 set in January. Its monthly trading volume also rebounded to a record $3.02 billion last month, after hovering around or below $1 billion from February to August. The platform opened 38,270 new markets in October, almost three times the number in August. Polymarket's trading volume, number of active traders, and number of new markets all reached record highs in October. Kalshi's trading volume even surpassed Polymarket in October, reaching $4.4 billion.

Furthermore, following the shift in US regulation and the acquisition of regulated entities, the path to compliance reverting to the US is becoming increasingly clear. These events collectively demonstrate that the market for information derivatives centered on prediction has genuine, robust, and mainstream-recognized demand.

From an application spillover perspective, prediction markets can be considered a universal risk hedging and governance module. Enterprises can hedge operational risks by assessing the probability of policy implementation; DAOs can use conditional markets to link proposals with KPIs; and media and platforms can use probabilistic narratives as a new layer for information presentation. The integration of information portals like Google and Perplexity with prediction platforms is accelerating this era where probability is the interface.

The investor dilemma amid a booming sector: assets available but not investable.

When a sector enters its early stages of explosive growth, ordinary investors typically ask two questions: First, is the demand genuine? Second, how can they share in the growth? We've already seen the answer to the former; however, the latter has long presented an awkward reality in predicting sectors: top-performing products are available but not necessarily investable.

Take Polymarket as an example. Its official statement initially indicated that the project had no token and announced no airdrop or TGE plans. Although Polymarket's Chief Marketing Officer, Matthew Modabber, recently confirmed the POLY token and airdrop plan, and founder Shayne Coplan also revealed plans to launch the POLY token earlier in October, this still means that for investors who didn't participate deeply in Polymarket's early stages, the most lucrative and asymmetrical period of initial gains has already been exhausted. Now, unless you personally participate in every event market, it's difficult to achieve sector-level beta exposure and long-term returns. For investors hoping to hold sector growth exponentially, such targets are extremely scarce.

More broadly, regulated event contract platforms like Kalshi also lack native crypto tokens; while other on-chain prediction applications or tools either lack the scale and network effect to act as industry indices, or are more like single-function tools, unable to bear the value attribution of a sector. The result is that demand is booming at the application layer, but there is a structural gap at the investment layer with no tokens available for investment.

From Pump.fun and Virtuals, let's look at Polymarket and DeAgent AI.

Looking back at the meme sector in 2024, one of the most representative phenomena was the breakout success of Pump.fun. Its extremely low barrier to entry and standardized curve issuance mechanism ignited the zero-to-one creation of on-chain tokens. In its early, rapid growth phase, the platform itself had no native token; users could only share in the prosperity by participating in individual stock-like games with each meme. Subsequently, the market saw the emergence of Virtuals (VIRTUAL), a tokenized carrier that could exponentially quantify this ecosystem-level popularity. By binding key paths such as creation, trading, and LP pairing within the ecosystem to the platform token, VIRTUAL made holding VIRTUAL almost equivalent to holding the growth index of the entire Agent/Meme ecosystem, thus absorbing the premium released by Pump.fun in terms of narrative and fundamentals.

Pump.fun launched its platform token PUMP in mid-to-late 2025, but the timing was later, and its value capture logic was out of sync with the early ecosystem boom. Historical experience tells us that when the application layer explodes first and there is a lack of index assets, the infrastructure projects that first provide both products and tokens often outperform the industry average in value revaluation.

Returning to the emerging prediction market sector, DeAgent AI plays precisely this infrastructure role. DeAgentAI is an AI agent infrastructure covering the Sui, BSC, and BTC ecosystems, empowering AI agents to achieve trustless autonomous decision-making on-chain. It aims to address the three major challenges facing AI in distributed environments: identity authentication, continuity assurance, and consensus mechanisms, building a trustworthy AI agent ecosystem.

DeAgent AI has built a core protocol around prediction markets and DeFi scenarios, centered on AI oracles and a multi-agent execution network. One end connects to real-world and on-chain data, standardizing complex judgments, decisions, and signal production into verifiable oracle outputs. The other end connects these outputs to trading, governance, and derivatives design through an agent network, thus becoming the information and value hub of the entire sector.

This is precisely why this mirror image is being replayed in the prediction market sector today. Polymarket corresponds to Pump.fun of yesteryear (a leading product but lacking investable tokens for a long time), while DeAgent AI (AIA) plays the role of a virtuals-like value container. It provides the key infrastructure modules missing from the prediction market (AI oracles and agent execution networks), and also provides the publicly tradable token AIA as an anchor for the indexation of the sector, allowing investors to indirectly share in the medium- to long-term growth of the entire prediction sector by holding AIA.

How DeAgent AI Becomes a Value Container for Predicting Tracks

The core of DeAgentAI's technical framework lies in addressing the three fundamental challenges of continuity, identity, and consensus faced by decentralized AI agents running on the blockchain. Through a state system combining hot and long-term memory, and on-chain state snapshots, agents are not reset across multiple chains and tasks, ensuring a complete and traceable lifecycle for their behavior and decisions. On-chain unique identities + DID and hierarchical authorization mechanisms ensure that each agent's identity is unforgeable. Furthermore, minimum entropy decision-making and validator consensus converge the chaotic outputs of multiple models into a deterministic, scalable result. Building upon this, the A2A protocol handles standardized collaboration between agents, while the MPC execution layer safeguards the privacy and security of sensitive operations. Ultimately, identity, security, decision-making, and collaboration are integrated into a verifiable and scalable decentralized AI agent infrastructure.

The dual-track deployment of AlphaX and CorrAI

At the application layer, AlphaX and CorrAI are the most tangible manifestations of this infrastructure. AlphaX, the first AI model developed by its community based on the DeAgentAI feedback training mechanism, employs the Transformer architecture, Mixture-of-Experts (MoE) technology, and Human Feedback Reinforcement Learning (RHF) mechanism, focusing on improving the accuracy of cryptocurrency price predictions. AlphaX predicts cryptocurrency price trends over 2–72 hours with an accuracy rate of 72.3%, achieving ROIs of +18.21% and +16.00% in live trading simulations in December 2024 and January 2025, respectively, with a win rate of around 90%, demonstrating the considerable practicality of AI predictions in real trading environments.

CorrAI is more like a no-code copilot for DeFi/quantitative users, helping users select strategy templates, adjust parameters, perform backtesting, and issue on-chain instructions, connecting the observed signals and the execution of strategies into a closed loop, and also bringing more real funds and behaviors into the DeAgent AI's agent network.

On the ecosystem side, AlphaX has accumulated a considerable number of users and interactions on public chains such as Sui and BNB through activities and integrations. With the support of multiple chains and various application scenarios, the DeAgent AI network has formed a production relationship with hundreds of millions of on-chain interactions and tens of millions of users. It is no longer an experimental project that remains in the white paper, but a real, running, and continuously invoked infrastructure.

From price feeds to subjective judgments: AI oracles

Traditional oracles primarily handle objective values like BTC/USD, relying on multi-node redundancy and data source aggregation to achieve consensus. However, once the question becomes a subjective/deterministic judgment (e.g., "Is ETH more likely to rise or fall this weekend?"), each node calls a large model, and the answers given are often inconsistent. It is also difficult to prove that a certain model was indeed called according to the agreement and that the result was obtained, and security and trust begin to fail.

DeAgent AI was designed from the outset with the DeAgentAI Oracle to address these subjective questions. Users submit questions in multiple-choice format and pay a service fee. Multiple AI agents in the network independently judge based on retrieval and reasoning, then vote. An on-chain contract aggregates the votes, selects the final result, and records it on the blockchain. This compresses the previously divergent AI output into a settleable, deterministic result. The question of whether to believe a particular node is replaced by verifying a publicly available voting and recording process. For the first time, AI making judgments becomes a public service that can be repeatedly invoked on-chain, making it highly suitable for scenarios such as prediction markets, governance decisions, and InfoFi. This component is currently undergoing internal testing.

In specific cases, DeAgent AI's agents have already been used to make judgments about real-world events. During the recent US federal government shutdown, the team built a decision tree model at the end of October based on market pricing from platforms such as Kalshi and Polymarket, combined with historical shutdown durations, the two-party game structure, and key time nodes. The final conclusion was that this round of shutdown was most likely to be forced to end between November 12-15 (or close to November 13-20), rather than the endless back-and-forth narrative commonly seen in market sentiment.

Concurrently, regarding the controversial topic of "whether Bitcoin has entered a bear market," DeAgent AI, by integrating signals such as on-chain data, ETF fund flows, macroeconomic policy shifts, and technical indicator divergences, judged that the current stage is closer to "a deep correction in the early stages of a bear market" rather than an accelerated bull market that has not yet ended, and based on this, provided key price levels and a risk monitoring framework.

These predictions and analyses focusing on specific issues demonstrate DeAgent AI oracles' ability to deconstruct and integrate subjective and complex problems. They also show that their outputs can be directly transformed into signals usable for market prediction and trading decisions, rather than just remaining at the demonstration level.

How AIA exponentializes track growth

From an investor's perspective, AIA's value capture logic lies in the fact that it serves as both a payment and settlement medium for the DeAgentAI Oracle and the Agent network, and a staked asset and governance credential for nodes and validators. As more prediction applications, governance modules, and DeFi strategies connect to this network, the number of requests, call frequency, and security requirements will translate into actual demand for AIA, naturally binding its value to the overall usage across the sector, rather than relying solely on one-off narrative hype.

More importantly, this value chain is closed-loop and predictable. As prediction applications like Polymarket expand market categories and introduce more complex subjective questions, they rely on AI oracles to make complex judgments; these calls directly reflect the increased demand for AI oracle infrastructure like DeAgent AI; and as the usage of the Oracle/Agent network increases, the demand and value of its associated token, AIA, as a payment, settlement, and staking asset, will also rise accordingly. In other words, if you believe the prediction market will continue to expand, it's hard not to simultaneously believe the demand for AI oracles will increase, and this will ultimately be reflected in the long-term pricing of AIA.

From an asset perspective, AIA simultaneously meets the criteria of "functionality" and "investability." On one hand, it corresponds to AI oracles and intelligent agent infrastructure addressing subjective problems, directly addressing the core pain points of the prediction market. On the other hand, it is itself a token asset that can be allocated in the public market. In contrast, prediction platforms such as Kalshi and Polymarket currently lack native tokens for investment, and while traditional price oracles have tokens, they serve the objective price feed track, not belonging to the same value chain as AI-driven subjective oracles. In the niche of AI oracles + tradable tokens, AIA is currently one of the few, if not the only, assets that simultaneously satisfies both usability and investability, thus having the potential to become the most direct indexation vehicle for the growth of the prediction sector.

How should one participate in predicting the market trend?

The current prediction market has clearly entered a phase where application stories are being told on the front page while value is gradually settling behind the scenes. Polymarket and Kalshi have proven the existence of this market with real trading volume. What can truly be priced in the long term is likely the layer that supports the operation of these applications, namely the AI oracles and intelligent agent networks responsible for judgment and settlement, as well as the functional tokens tied to them.

As predictive applications attempt to handle more complex and subjective judgments, they will inevitably generate higher and more frequent demands for AI oracles. This demand will ultimately translate into continued use of infrastructure like DeAgent AI. The functional tokens closely tied to this infrastructure for payments, settlements, and staking will also absorb corresponding value in this process. Therefore, the real question now is no longer whether to participate in this field, but rather how and at what level to participate.

A relatively clear approach is to use participation at the application layer and position size at the infrastructure layer. At the application layer, users can continue to use platforms like Polymarket as tools to acquire alpha, using position size to bet on specific events; at the infrastructure layer, a moderate allocation of AIA (AI Oracle Algorithm) is used to align with the longer-term goal of AI oracles becoming a standard feature of the prediction market. The former answers whether it's profitable in this particular instance, while the latter answers whether, as the market grows, it will benefit from the underlying infrastructure.

Of course, AIA is just one factor in the portfolio, not a substitute for risk control itself. A more prudent approach is to treat it as part of predicting the infrastructure index of the sector, giving this long-term logic a place and time within your own risk budget, and allowing the market to validate your judgment of this narrative.