AI + prediction market = future? Web3 is using Token mechanisms to solve the "Crisis of Confidence" of the information age.

Original author: KarenZ

Reprint: Daisy, Mars Finance

In 1971, psychologist and economist Herbert A. Simon first proposed the theory of attention economy, pointing out that in a world of information overload, human attention has become the most scarce resource.

Economist and USV managing partner Albert Wenger further reveals a fundamental shift in "The World After Capital": human civilization is undergoing a third leap—from the "capital scarcity" of the industrial age to the "attention scarcity" of the knowledge age.

Agricultural Revolution: Aimed at solving food scarcity issues, but led to land disputes;

Industrial Revolution: Committed to solving the land scarcity issue, but turned to resource competition and capital accumulation;

Digital Revolution: Competing for Attention.

The underlying driving force behind this transformation comes from two key characteristics of digital technology: the zero marginal cost of information replication and dissemination, and the universality of AI computation (though human attention cannot be replicated).

Whether it's Labubu's popularity in the trendy toy market or the live streaming sales by top influencers, both are largely about the competition for user and audience attention. However, in the traditional attention economy, users, fans, and consumers contribute their attention as "data fuel," while the excess profits are monopolized by platforms and scalpers. The InfoFi of the Web3 world attempts to overturn this model—by utilizing blockchain, token incentives, and AI technology to make the processes of information production, dissemination, and consumption transparent, aiming to return value to the participants.

This article will provide an in-depth introduction to the classification of the InfoFi project, the challenges it faces, and the trends for future development.

What is InfoFi?

InfoFi is a combination of Information + Finance, focusing on transforming difficult-to-quantify, abstract information into dynamic, quantifiable value carriers. This encompasses not only traditional prediction markets but also the distribution, speculation, or trading of information or abstract concepts such as attention, reputation, on-chain data or intelligence, personal insights, and narrative activity.

The core advantages of InfoFi are reflected in:

Value Redistribution Mechanism: Returning the value monopolized by platforms in the traditional attention economy to the true contributors. Through smart contracts and incentive mechanisms, allowing information producers, disseminators, and consumers to share the benefits.

Information Valuation Capability: Transforming abstract concepts such as attention, insights, reputation, and narrative activity into tradable digital assets, creating a trading market for the previously illiquid value of information.

Low barrier to entry: Users can participate in value distribution through content creation using only their social media accounts.

Innovation in incentive mechanisms: not only rewarding content creation but also including multiple aspects such as dissemination, interaction, and verification, allowing niche content and long-tail users to also receive rewards. High-quality content receives more rewards, encouraging the continuous production of high-quality information.

Cross-domain application potential: For example, the introduction of AI provides advantages such as content quality assessment and market optimization for InfoFi.

InfoFi Classification

InfoFi covers a variety of different application scenarios and models, which can be mainly divided into the following categories:

Prediction Market

Prediction markets, as a core component of InfoFi, are a mechanism for forecasting the outcomes of future events through collective intelligence. Participants express their expectations about future events (such as election or policy outcomes, sports events, economic forecasts, price expectations, product release dates, etc.) by buying and selling "shares" linked to specific event outcomes, with market prices reflecting the collective expectations of the crowd regarding the event outcomes. Polymarket is a representative application promoting the concept of InfoFi.

Vitalik has always been a loyal supporter of the prediction market Polymarket. In his article "From prediction markets to info finance" published in November 2024, he stated, "Prediction markets have the potential to create better applications in social media, science, news, governance, and other areas. I call these types of markets info finance." Vitalik also pointed out the dual nature of Polymarket: one is a betting site for participants, and the other is a news site for everyone else.

In the framework of InfoFi, prediction markets are not just tools for speculation, but platforms for uncovering and revealing real information through financial incentive mechanisms. This mechanism leverages market efficiency, encouraging participants to provide accurate information, as correct predictions yield economic rewards while incorrect predictions may result in losses. Musk himself retweeted data stating "Trump leads with a 51% approval rating on Polymarket" one month before the 2024 U.S. election and commented: "Due to the involvement of real money, this data is more accurate than traditional polls."

The representative platforms of prediction markets include:

Polymarket: The largest decentralized prediction market, Polymarket is built on the Polygon network and uses USDC stablecoin as the trading medium. Users can make predictions on events such as political elections, economics, entertainment, and whether products will be launched.

Kalshi: is a prediction market platform in the U.S. fully regulated by the CFTC, supporting deposits in USDC, BTC, WLD, SOL, XRP, and RLUSD through a partnership with the cryptocurrency and stablecoin infrastructure provider Zero Hash, but settled in fiat currency. Kalshi focuses on Event Contracts, allowing users to trade the outcomes of political, economic, and financial events. Due to regulatory compliance, Kalshi has a unique advantage in the U.S. market.

Mouth Licking InfoFi (Yap-to-Earn)

"Zui Lu" is a humorous term used in the Chinese cryptocurrency community to refer to Yap-to-Earn, which means earning rewards by expressing opinions and sharing content. The core concept of Yap-to-Earn is to encourage users to post high-quality, cryptocurrency-related posts or comments on social media platforms. Most of the time, AI algorithms evaluate the quantity, quality, interaction, and depth of the content to allocate points or token rewards. This model differs from traditional on-chain activities (such as trading or staking) and focuses more on users' contributions and influence within the community.

Characteristics of "Zui Lu":

No on-chain transactions or high capital are required; you only need an X account to participate.

Enhance the activity of the project community through rewarding valuable discussions.

AI algorithms reduce human intervention, filtering out bots and low-quality content, ensuring a more transparent reward distribution.

Points may be converted into token airdrops or ecological privileges, and early participants may receive higher returns.

Current mainstream mouth-play projects or projects that support mouth-play include:

Kaito AI: is the representative platform for Yap-to-Earn, having collaborated with multiple projects to evaluate the quantity, quality, interactivity, and depth of users' crypto-related content published on X through AI algorithms, rewarding Yap points for users to compete for token airdrops on the leaderboard.

In this way, creators can not only effectively prove their influence and content value through Yaps, but also attract precise high-quality attention; ordinary users can efficiently discover quality content and KOLs through the Yaps system; while project parties achieve the dual goals of accurately reaching target users and expanding brand influence, forming a healthy ecological cycle of mutual benefit.

Kaito AI has distributed tokens worth over $90 million to various communities (excluding Kaito's own airdrops), with over 200,000 active Yappers each month.

Source:

Cookie.fun: Cookie tracks the mindshare, interactions, and on-chain data of AI agents, generating a comprehensive market overview, and also tracks the mindshare and sentiment of cryptocurrency projects. Cookie Snaps has a built-in rewards and airdrop system that provides rewards to Cookie creators who contribute to project attention.

Cookie has collaborated with three projects to launch the Snaps event, namely Spark, Sapien, and OpenLedger. Among them, the number of participants in the Spark event exceeded 16,000, while the participation numbers for the other two projects were 7,930 and 6,810 respectively.

Virtuals: Virtuals is not a platform focused on Yap-to-Earn, but rather an AI agent launch platform. However, in mid-April, it launched a new launch mechanism called Genesis Launch on Base. One way to earn points to participate in the launch includes Yap-to-Earn (supported by Kaito).

The top AI agency projects with high subscription rates on Virtuals, source:

Loud: As a "Attention Value Experiment" within the Kaito AI ecosystem, Loud occupied over 70% of the Kaito Attention Ranking through the Yap-to-Earn campaign before the official release of its token via the Initial Attention Offering (IAO) at the end of May 2025. The LOUD operating mechanism also revolves around the "attention economy," with transaction fees collected after trading primarily distributed in SOL to the top 25 users on the attention ranking.

Wallchain Quacks: Wallchain is a programmatic AttentionFi project based on Solana, supported by AllianceDAO. Wallchain X Score assesses the overall influence of users, while Wallchain Quacks rewards high-quality content and valuable interactions. Currently, Wallchain Quacks custom LLM evaluates creator content daily, and valuable, insightful content creators will receive Quacks rewards.

Mouth Licking + Tasks / On-chain Activities / Verification: Multi-dimensional Contribution Value Realization

Some projects also evaluate users' multidimensional contributions by combining content contributions with on-chain activities (such as transactions, staking, and NFT minting) or tasks.

Galxe Starboard: Galxe is a Web3 growth platform, and its newly launched Galxe Starboard is dedicated to rewarding genuine contributions in off-chain and on-chain actions. Projects can define multiple layers of contribution, where the importance lies not just in how many tweets were sent, but in the value brought to the entire project, including post engagement, sentiment, virality, interaction with dApps, holding tokens, minting NFTs, or completing on-chain tasks.

Mirra: Mirra is a decentralized AI model trained on community-curated data that learns from the real-time contributions of Web3 users. Specifically, creators publish high-quality content on X, which acts as a submission of AI verification data; Scouts identify high-value content on X and tag @MirraTerminal in replies to submit insights, determining what content the AI learns from and helping to shape intelligent AI.

Reputation-based InfoFi

Ethos is an on-chain reputation protocol that is entirely based on open protocols and on-chain records, combined with Social Proof of Stake (Social PoS), to generate a Credibility Score through decentralized mechanisms, ensuring the reliability, decentralization, and resistance to Sybil attacks of its reputation system. Currently, Ethos adopts a strict invitation system. The core function of Ethos is to generate a Credibility Score, a quantitative indicator of a user's trustworthiness on-chain. The scoring is based on the following on-chain activities and social interactions: a commenting mechanism (with cumulative utility) and a guarantee mechanism (staking Ethereum to endorse other users).

Ethos also launched a reputation market that allows users to speculate on the reputation of individuals, companies, DAOs, and even AI entities by buying and selling "trust votes" and "distrust votes," effectively going long or short on reputation.

GiveRep: Primarily built on Sui, it aims to transform users' social influence and community engagement on the X platform into quantifiable on-chain reputation through their activities, incentivizing users to participate with rewards. Commenting on creator posts by mentioning the official GiveRep Twitter will earn both the commenter and the creator a reputation point. To limit abuse, GiveRep restricts users to a maximum of 3 such comment mentions per day (including 3), while creators can receive unlimited points daily. Comments from Sui ecosystem projects and ambassadors will earn more points.

Attention Market / Prediction

Noise: is a trend discovery and trading platform based on MegaETH, currently requiring an invitation code for access. Users can long or short the project's attention.

Upside: Upside is a social prediction market (investors include Arthur Hayes) that rewards the discovery, sharing, and prediction of valuable content and links, creating a dynamic market through a like mechanism. Earnings are proportionally distributed to voters, creators, and curators. To prevent manipulation of the prediction pool, the weight of likes will decrease in the last 5 minutes of each round.

YAPYO: An attention market infrastructure in the Arbitrum ecosystem. YAPYO indicates that the rewards in its coordination mechanism are not only profits but also enduring influence.

Trends: You can tokenize X posts to become a trend on the joint curve, referred to as "Trend it." Creators are eligible to receive 20% of the joint curve trading fee for each trend.

Token-gated content access: filtering noise

Backroom: Creators can launch tokenized spaces that provide curated content such as market insights, Alpha, and analysis without the need for management or social pressure; users can unlock low-noise, high-value information by purchasing on-chain Keys tied to each creator's space. Keys are not just for access—they are tradable assets with dynamically priced curves driven by demand. At the same time, AI will process chat data and signals into actionable insights.

Xeet: A new protocol on the Abstract network that has not been fully launched yet, but a referral program has been introduced, inviting KOLs to earn reward points. Xeet founder @Pons_ETH mocked that InfoFi has evolved into NoiseFi, stating, "It's time to reduce the noise and enhance the signal." The currently available information is that Xeet will integrate with Ethos score, and besides that, Xeet has not disclosed more information.

Data Insights Class InfoFi

Arkham Intel Exchange: Arkham is an on-chain data query tool, intelligence trading platform, and exchange. Arkham Intel Intel Exchange is a decentralized intelligence trading platform where "on-chain detectives" can earn bounties.

InfoFi Dilemma

Prediction Market

Regulation and Compliance: Prediction markets may be viewed as akin to binary options and gambling markets, facing regulatory pressure. For example, Polymarket was deemed to be operating illegally by the CFTC in the United States for not being registered as a designated contract market (DCM) or swap execution facility (SEF), leading to a fine of $1.4 million in 2022 and a requirement to block U.S. users. The investigation by the U.S. Department of Justice and the FBI in 2024 further highlighted its regulatory predicament.

Insider Trading and Fairness: Predictive markets may be affected by insider information. Large funds may distort prices in the short term. Designing fair rules and mechanisms is one of the key challenges for InfoFi predictive markets.

Liquidity and Participation: The effectiveness of prediction markets depends on sufficient participants and liquidity. Prediction markets often face the "long tail liquidity problem" on niche topics, where the lack of participants leads to unreliable market information. The introduction of AI agents may partially address this issue, but further optimization is still needed.

Oracle Design: Polymarket has previously encountered oracle manipulation attacks, resulting in significant losses for users who bet on the correct outcome. In February 2025, UMA, Polymarket, and EigenLayer announced they are collaborating to research the construction of predictive market oracles. Some research ideas include developing an oracle that can support multiple tokens to resolve disputes, with other features under investigation including dynamic binding, AI agent integration, and enhanced security against bribery attacks.

Mouth licking

The information noise has intensified, and AI content advertising accounts are rampant, obscuring the real signals. Users find it difficult to filter valuable information from the massive amount of content, community trust is declining, and the marketing effectiveness for project parties is diminished. According to KOL Crypto Brave (@cryptobraveHQ), "Several project owners have complained that after spending 150,000 USDT service fee on Kaito, allocating 0.5%-1% of tokens to KOLs, a large portion of the participants are AI content advertising accounts. Project parties want to attract top KOLs and ICT to participate, but they have to pay extra, and then Kaito contacts the top KOLs to get involved."

Most mouth-lifting projects lack public explanations of their algorithms for assessing content quality, interactivity, and depth, raising user concerns about the fairness of point distribution. If the algorithm favors specific accounts (such as influencers or matrix accounts), it may lead to the loss of quality creators. Recently, Kaito has made some new upgrades to the algorithm based on community feedback, focusing on prioritizing quality over quantity by default, ensuring that posts without project insights and comments do not receive attention, and further cracking down on interaction manipulation and group spamming behaviors.

The Matthew effect of profit distribution: In most cases, projects and KOLs achieve a win-win situation, but tail content creators and interactive retail investors still face the dilemma of low earnings and intense competition. Kaito founder Yu Hu stated on June 8 that "of the approximately 1 million registered users on Kaito, fewer than 30,000 users have earned yaps, which is less than 3%. The next growth phase for the network is to maximize conversion rates." Additionally, poorly managed airdrop expectations can lead to community dissatisfaction. Magic Newton is a relatively successful case of mouth-to-mouth promotion on Kaito AI, with Kaito ecosystem recommendations accounting for 1/3 of all Newton validation agents. Mouth-to-mouth users have made a significant profit, but they also face accusations of being unfriendly to retail investors. In contrast, Humanity has been directly accused by the community of "betraying users" and "extreme anti-loyalty," and this distribution imbalance has triggered a crisis of trust.

The initial phase of the mouth-licking activity attracted user participation, but attention experienced a cliff-like drop after the rewards were distributed, lacking sustainability. LOUD's market capitalization on the day of launch once approached 30 million USD, but currently, it has dwindled to less than 600,000 USD.

Attention does not equal market cap share.

reputation

Reputation InfoFi projects like Ethos adopt an invitation-based system to control user quality and reduce the risk of female witch attacks. However, this mechanism raises the participation threshold, limiting the onboarding of new users and making it difficult to create widespread network effects.

Malicious operation risk.

The issue of cross-platform recognition of reputation scores arises, as different protocols have rating systems that are difficult to interoperate, creating information silos.

InfoFi Trend

Prediction Market

The combination of AI and prediction markets: AI can significantly enhance the efficiency of prediction markets. For example, AI can provide more accurate predictions in complex scenarios by analyzing vast amounts of data; it can also explore AI agents to solve long-tail problems.

The combination of social media and prediction markets: Prediction markets have the potential to become the core infrastructure of the information economy in the future. On June 6, X officially announced a partnership with Polymarket, which becomes X's official prediction market partner. Polymarket's founder and CEO Shayne Coplan stated, "Combining Polymarket's accurate, fair, and real-time prediction market probabilities with Grok's analytics and X's real-time insights will be able to provide contextual, data-driven insights instantly to millions of Polymarket users worldwide."

Decentralized Governance: Prediction markets can be applied to the governance of DAOs, companies, and even societies, known as "Futarchy." Vitalik stated in 2014 that Futarchy is a governance model proposed by economist Robin Hanson, with the core idea being "vote values, bet beliefs." The operation works as follows: the community votes to determine a metric for success (such as GDP, company stock prices, etc.); for specific policy proposals, two prediction markets are created (for and against). Participants trade these two tokens, and the price reflects the market's expectation of whether the policy will optimize the target; ultimately, the policy with the higher average price is chosen, and token returns are settled based on actual results. The advantage of Futarchy is its reliance on data rather than political propaganda, personal charisma, or promotion.

Content and news tools for everyone.

Mouth Licking + Reputation Type InfoFi

Introducing social graph and semantic understanding technologies to enhance the accuracy of AI's content value assessment, ultimately aiming for high-quality content.

Incentivize high-quality long-tail creators.

Add reduction or penalty mechanisms.

Release of InfoFi LLM for Web3.

Multi-dimensional assessment of contributions.

Reputation-based InfoFi combines with DeFi, using reputation scores as the credit basis for lending and staking.

The tokenization of abstract assets such as attention, reputation, and trends will give rise to more types of derivatives.

Not just based on X social platform.

The integration with more social platforms and news media drives the formation of a tool for attention and Alpha discovery aimed at everyone.

Data Insight Type InfoFi

The combination of data analysis charts and creator insights, while also adding incentive mechanisms related to creation, distribution, etc.

The combination of data analysis charts and AI analysis.

Summary

The core contradiction of the digital age is the divide between attention creators and value holders. This divide is the driving force of the Web3 InfoFi revolution.

The core contradiction of InfoFi lies in the inability to balance information value and participation incentives, which may lead to a repeat of the SocialFi "high opening and low going" scenario. The key to InfoFi is to establish a "trinity" balance mechanism among information mining, user participation, and value return, thus driving the formation of a better knowledge-sharing and collective decision-making infrastructure. This not only requires the technical level to achieve attention quantification but also ensures that ordinary participants can obtain reasonable returns from information dissemination in the mechanism design, avoiding severe imbalances in value distribution.

More importantly, the revolution of InfoFi requires a joint effort from both top-down and bottom-up approaches to truly achieve fairness and efficiency in the attention economy. Otherwise, the Matthew effect of the profit pyramid will reduce InfoFi to a gold mining game for a few, which goes against the original intention of "universal attention value."