How does DeAgent AI become the value hub of the prediction track outside of Polymarket?

Written by: ChandlerZ, Foresight News

If human society has had a curiosity and bet on the future since ancient times, then the crypto-native prediction markets are transforming this ancient demand into a public good that is quantifiable, liquid, and reusable. Over the past decade, the democratization of information has been achieved through the internet; in the Web3 and crypto space, values and beliefs are also being tokenized and priced, forming a more verifiable and incentive-compatible value democratization. The addition of AI has expanded the boundaries of prediction from simple price feeding to more complex judgments and rulings, giving predictions the meaning of infrastructure, eliminating speculative interpretations. Prediction markets serve as the foundational information cornerstone for governance, hedging, and resource allocation. Google's integration of Polymarket and Kalshi's market probabilities into Google Finance starting in November 2025 marks the entry of prediction data into the public layer with hundreds of millions of users, which is both an endorsement of the industry and a signal of incremental demand.

Why Prediction Markets are a Battleground for Web3

The essence of prediction markets is to aggregate the tacit knowledge dispersed in individual minds into public probabilities through prices. This idea can be traced back to Robin Hanson's Futarchy (governance by betting), where the value goals are determined by voting, and factual judgments are priced by the market, with prediction markets set as the primary mechanism for information aggregation. Research in academia has also shown that prediction markets outperform simple polls in many scenarios regarding event outcomes, especially in dynamic updates and incentive constraints.

When shifting the perspective from theoretical reasoning back to the real market, you will find that this mechanism of aggregating cognition through prices is being voted on by funds and users in 2024-2025. Prediction platforms represented by Polymarket and Kalshi have seen daily trading volumes approaching or even exceeding $100 million, with cumulative transaction volumes surging to hundreds of billions of dollars, marking the transition of prediction markets from niche experiments to full-blown explosions. Data shows that Polymarket reached a historical high of 477,850 monthly active traders in October, surpassing the previous record of 462,600 set in January. Its monthly trading volume rebounded to a record $3.02 billion last month, after hovering around or below $1 billion from February to August. The number of new markets opened on the platform in October reached 38,270, nearly three times that of August. Polymarket set historical highs in trading volume, active traders, and new market openings in October. Kalshi's trading volume in October even surpassed Polymarket, reaching $4.4 billion.

In addition, with the shift in U.S. regulation and the mergers of regulated entities, the path for compliance to return to the U.S. is becoming clearer. This series of events collectively indicates that the information derivatives market centered around predictions has a real, strong demand recognized by mainstream entry points.

From the perspective of application spillover, prediction markets can be seen as a universal risk hedging and governance module. Enterprises can hedge operational risks against the probability of policy implementation, DAOs can bind proposals and KPIs with conditional markets, and media and platforms can use probability narratives as a new layer of information display. The integration of information entry points like Google and Perplexity with prediction platforms is accelerating this era where probability equals interface.

The Investor Dilemma Under Track Prosperity: Usable but Not Investable

When a track enters early-stage explosion, ordinary investors usually ask two questions. One is whether the demand is real, and the other is how to share in the growth. We have already seen the answer to the former; however, the latter has long been trapped in an awkward reality in the prediction track: leading products are usable but not investable.

Taking Polymarket as an example, its officials once stated that the project has no token and has not announced any airdrop or TGE plans. Although recently, Polymarket's Chief Marketing Officer Matthew Modabber confirmed the POLY token and airdrop plans. Earlier in October, the company's founder Shayne Coplan also revealed that they would launch the POLY token. But this still means that for investors who did not deeply participate in Polymarket early on, the most lucrative and asymmetric original dividend period has essentially been consumed in advance. Now, unless you personally participate in every event market, it is difficult to gain track-level beta exposure and align long-term returns. For investors hoping to hold track growth in an index-like manner, the targets are extremely scarce.

More broadly, regulated event contract platforms like Kalshi also do not have crypto-native tokens; other on-chain prediction applications or tools either lack the scale and network effects to act as industry indices or resemble single-function tools that cannot carry track-level value attribution. The result is that demand is blooming fiercely at the application level, while the investment layer has a structural gap with no tokens to invest in.

From Pump.fun and Virtuals: Observing Polymarket and DeAgent AI

Looking back at the Meme track of 2024, one of the most representative phenomena is the breakout of Pump.fun, with its extremely low barriers to entry and standardized curve issuance mechanism igniting the zero-to-one creation on-chain. In its early explosive phase, the platform itself had no native token, and users could only share in the prosperity by participating in each meme's individual stock-like bets. Subsequently, the market saw the emergence of a token vehicle, Virtuals (VIRTUAL), that could index this ecological heat. VIRTUAL binds key paths such as creation, trading, and LP pairing within the ecosystem to the platform token, making holding VIRTUAL akin to holding a growth index of the entire Agent/Meme ecosystem, thereby capturing the premium released by Pump.fun in both narrative and fundamentals.

Pump.fun launched its platform token PUMP in the latter half of 2025, but the timing was later, and its value capture logic misaligned with the earlier ecological explosion. Historical experience tells us that when the application layer explodes first without an index asset, the infrastructure projects that provide both existing products and tokens often outperform the average of the track in value reassessment.

Returning to the emerging prediction market track, DeAgent AI plays such an infrastructural role. DeAgent AI 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 three major challenges faced by AI in distributed environments: identity authentication, continuity assurance, and consensus mechanisms, thereby building a trustworthy AI agent ecosystem.

DeAgent AI has constructed a set of underlying protocols centered around prediction markets and DeFi scenarios, with AI oracles and multi-agent execution networks at its core. One end connects real-world and on-chain data, standardizing complex judgments, rulings, and signal production into verifiable oracle outputs, while the other end integrates these outputs into trading, governance, and derivatives design through an agent network, thus becoming the information and value hub of the entire track.

Because of this, this framework is currently being mirrored in the prediction market track. Polymarket corresponds to the earlier Pump.fun (product leader but long lacking investable tokens), while DeAgent AI (AIA) plays the role of a value container similar to Virtuals. It not only provides the key infrastructure modules missing in prediction markets (AI oracles and agent execution networks) but also offers a publicly tradable token AIA as an anchor for track indexing, allowing investors to indirectly share in the medium to long-term growth of the entire prediction track by holding AIA.

How DeAgent AI Becomes the Value Container of the Prediction Track

The technical framework of DeAgent AI focuses on solving the three fundamental challenges of continuity, identity, and consensus faced by decentralized AI agents running on-chain. Through a state system combining hot memory and long-term memory, as well as on-chain state snapshots, agents will not be reset in multi-chain and multi-task scenarios, and their behaviors and decisions will have a complete, traceable lifecycle; using unique on-chain identities + DID and hierarchical authorization mechanisms ensures that each agent's identity is non-falsifiable; and employing Minimum Entropy Decision and validator consensus converges the chaotic outputs of multiple models into verifiable deterministic results. On this basis, the A2A protocol is responsible for standardized collaboration between agents, while the MPC execution layer ensures the privacy and security of sensitive operations, ultimately integrating identity, security, decision-making, and collaboration into a verifiable and scalable decentralized AI agent infrastructure.

The Dual Landing of AlphaX and CorrAI

At the application layer, AlphaX and CorrAI are the most intuitive realizations of this infrastructure. AlphaX is the first AI model developed based on the feedback training mechanism of DeAgent AI, incubated by its community, utilizing 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 for 2-72 hours, achieving an accuracy rate of 72.3%, and recorded +18.21% and +16.00% ROI in real 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, on the other hand, resembles a no-code Copilot for DeFi/quantitative users, helping users select strategy templates, adjust parameters, backtest, and issue on-chain commands, closing the loop between seeing signals and executing strategies, while also bringing more real funds and behaviors into the DeAgent AI agent network.

On the ecological side, AlphaX has already accumulated a considerable number of users and interactions through activities and integrations on public chains like Sui and BNB. Coupled with multiple chains and various application scenarios, the overall network of DeAgent AI has formed a production relationship with hundreds of millions of on-chain interactions and tens of millions of users, no longer remaining an experimental project confined to white papers, but rather a real, continuously utilized infrastructure.

From Price Feeding to Subjective Judgment AI Oracles

Traditional oracles primarily handle objective values like BTC/USD, achieving consensus through multi-node redundancy and data source aggregation; once the issue becomes subjective/non-deterministic judgments (e.g., "Is ETH more likely to rise or fall this weekend?"), the nodes independently call large models, and the answers they provide often do not align, making it difficult to verify that a certain model was indeed called and yielded that result, leading to failures in safety and trust.

From the outset, DeAgent AI designed the DeAgentAI Oracle to address such subjective issues. Users submit questions in the form of multiple-choice questions and pay a service fee, with multiple AI agents in the network independently judging based on retrieval + reasoning and then voting. The on-chain contract aggregates the votes, selects the final result, and records it on-chain. In this way, the originally divergent AI outputs are compressed into verifiable deterministic results, replacing the belief in a certain node with the verification of a publicly available voting and recording process, making the act of AI judgment a public service that can be repeatedly called on-chain, highly suitable for prediction markets, governance decisions, and InfoFi scenarios. This component is currently undergoing internal testing.

In specific cases, DeAgent AI's agents have already been used to provide judgments around real-world events. Recently, during the U.S. federal government shutdown, the team constructed a decision tree model at the end of October based on market pricing from platforms like Kalshi and Polymarket, combined with historical shutdown durations, bipartisan negotiation structures, and key time nodes, concluding that this round of shutdown is most likely to end between November 12-15 (or in the vicinity of November 13-20), rather than the commonly seen narrative of endless tug-of-war in market sentiment.

At the same time, regarding the controversial topic of "Has Bitcoin entered a bear market?", DeAgent AI assessed that the current phase is closer to "early deep adjustments of a bear market," rather than an ongoing accelerated bull market, based on signals from on-chain data, ETF fund flows, macro policy shifts, and technical indicator divergences, and provided key price levels and risk monitoring frameworks accordingly.

This type of prediction and assessment around specific topics not only demonstrates DeAgent AI's oracle's ability to decompose and integrate subjective, complex issues but also indicates that its outputs can directly translate into usable signals for prediction markets and trading decisions, rather than merely remaining at the demonstration level.

How AIA Indexes Track Growth

Returning to the investor's perspective, the value capture logic of AIA lies in its dual role as both the payment and settlement medium for DeAgentAI Oracle and Agent networks, as well as the staking asset and governance certificate for nodes and validators. As more prediction applications, governance modules, and DeFi strategies integrate into this network, the number of requests, call frequency, and security needs will translate into actual demand for AIA, binding its value to the overall usage of the track, rather than relying solely on one-time narrative hype.

More critically, this value chain is both closed-loop and deducible. As prediction applications expand market categories and introduce more complex subjective issues, they will inevitably rely on AI oracles for complex judgments; these calls will directly reflect the growing demand for AI oracle infrastructure like DeAgent AI; and as the usage of the Oracle/Agent network increases, the function token AIA, which is bound to it, will also see its demand and value rise accordingly. In other words, if you believe that prediction markets will continue to expand, it is hard not to simultaneously believe that the demand for AI oracles will grow, which will ultimately be reflected in the long-term pricing of AIA.

From an asset attribute perspective, AIA meets both the "functional" and "investable" criteria. On one hand, it corresponds to the AI oracle and agent infrastructure aimed at subjective issues, directly addressing the core pain points of prediction markets; on the other hand, it is a token asset that can be allocated in the public market. In comparison, platforms like Kalshi and Polymarket still do not have native tokens to invest in, while traditional price oracles may have tokens but serve the objective price feeding track, which is not 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, targets that can simultaneously meet the criteria of being usable and investable, thus having the opportunity to become the most direct indexation vehicle for growth in the prediction track.

How to Participate in the Prediction Track?

The current prediction track has clearly entered a stage where application stories are told upfront, and value gradually sinks below the surface. Polymarket and Kalshi have proven the existence of the track with real trading volumes, while what can truly be priced long-term is likely the layer that supports these applications, namely the AI oracles, agent networks responsible for judgments and settlements, and the function tokens bound to them.

As prediction applications attempt to carry more complex and subjective judgments, they will inevitably generate higher and more frequent demands for AI oracles; this demand will ultimately settle into the sustained use of infrastructure like DeAgent AI; and the function tokens closely tied to the payment, settlement, and staking of this infrastructure will also carry corresponding value in this process. Therefore, what truly needs to be considered next is not whether to participate in this track, but rather how to participate in this track and at what level.

A relatively clear approach is to engage at the application layer with participation and at the infrastructure layer with positions. At the application layer, users can continue to use platforms like Polymarket as tools to gain Alpha, betting on specific events with their positions; at the infrastructure layer, they can align with the long-term proposition of AI oracles becoming standard in prediction markets by moderately allocating AIA. The former answers whether one can make money in this round, while the latter addresses whether one has been lifted alongside the underlying when this track grows.

Of course, AIA is just one factor in the mix, not a substitute for risk control itself. A more prudent approach is to view it as part of the prediction track infrastructure index, giving this long-term logic a position and time within one's risk budget, allowing the market to validate the judgment of this narrative.