Multicoin: Attention Assets and Prediction Market

Author: Eli Qian, Investment Partner at Multicoin Capital; Translator: Jinse Finance xiaozou

For the sake of simplicity, we can roughly divide assets into two categories:

1 Cash Flow Assets - primarily stocks and bonds. These assets generate the cash flow that investors value.

2. Supply and Demand Assets - Mainly applicable to commodities and foreign exchange. Prices fluctuate with supply and demand relationships.

In recent years, the cryptocurrency space has given rise to a new type of asset—assets measured by attention value. Currently, “attention assets” mainly manifest as user-generated assets, such as NFTs, creator tokens, and Memecoins. These assets serve as the Schelling point of cultural attention tides and reflect fluctuations in attention through price volatility.

Although Memecoins are culturally interesting, they still have financial shortcomings. Efficient attention assets should allow market participants to establish risk exposure based on their direct attention to specific things. Through this mechanism, participants will be willing to trade assets they believe are mispriced, and the market will collectively form prices that reflect attention expectations.

We believe that through reasonable architectural design, attention assets are expected to be upgraded to a formal asset class. To promote this idea, this article proposes the concept of “Attention Oracle”—this new type of oracle architecture can support the creation of “Attention Perpetual Contracts,” allowing traders to take long and short positions based on the attention given to cultural symbols.

In short, the attention oracle collects binary prediction market data on specific topics, combining price, liquidity, and time dimensions to construct a weighted composite index to capture changes in attention. To ensure effective operation, it is crucial to carefully select the underlying markets to represent the real-world genuine attention input. Using prediction markets as a data source inherently incorporates manipulation costs—malicious traders need to invest capital to influence the index, which theoretically can suppress tampering behavior.

1**, Why Attention Should Be Paid to Perpetual Contracts**

User Generated Assets (UGAs) have achieved product-market fit in the realm of pure speculation and excel at tracking the attention of things that start from zero, such as emerging online trends and memes.

The core value of UGAs lies in creating assets for subjects that cannot exist through traditional financial channels. The traditional asset issuance process is slow, costly, and has high regulatory barriers, which greatly limits the range of subjects. Attention assets must maintain the speed of the internet to match the evolution of global thought. The combination of permissionless token issuance, bonding curves, and decentralized exchanges allows anyone to create assets for free, guide liquidity, and open trading to the world.

It has been observed that the price of UGAs typically starts from zero. This is not a flaw but a feature—when you create a new meme, its initial level of attention is zero. Entering at a low position is intuitive and allows those skilled at early trend detection to monetize low-cost underlying assets. However, this also makes it difficult for UGAs to effectively track existing things that already have high levels of attention.

For example, suppose you are optimistic about the attention on LeBron James and want to go long. While meme coins can be created, there are already dozens of existing LeBron tokens; how do you choose? Moreover, new coins need to start from scratch, and as a global top celebrity, his attention should ideally be high, making it unlikely for a short-term explosive increase of a hundred times. What if you want to short his attention? Meme coins find it even harder to support this operation.

So, what characteristics should assets with high attention targets possess? They must meet the following requirements:

• A two-way trading mechanism is required to support both long and short positions;
• Need to anchor real-world attention measurement benchmarks;
• Initial valuation should not start from zero.

If we step back and examine these requirements, we will find that perpetual contracts precisely meet the criteria: they allow for two-way operations, have an oracle pricing mechanism, and as derivatives, do not need to start from scratch. The real challenge lies in building the oracle system for attention perpetual contracts.

Some teams are already working to solve this problem, such as Noise. On this platform, traders can take long and short positions on the community sentiment shares of crypto projects like MegaETH and Monad. Noise uses Kaito as an oracle, generating numerical representations of topic popularity by aggregating social media and news data.

However, the existing design still has room for optimization. The core goal of the attention oracle is to collect attention-related data and process it through algorithms to output value indicators available for long and short trading.

The drawback of using social media as a data source is its susceptibility to manipulation — this confirms Goodhart’s Law: in adversarial markets, traders will attempt to manipulate pricing inputs. Kaito has had to redesign the leaderboard and spam filters to address this issue.

Moreover, social media is not a perfect measure of attention. Take Shohei Ohtani as an example: he has a global fan base that uses different social applications, and this data may not be fully captured by Kaito. If he wins the World Series again, his visibility will further increase, but the number of fans and mentions may not grow linearly.

2**, Attention Oracle: Market-based Solutions**

Returning to the case of LeBron James, suppose you want to trade on his attention. The first step in building a LeBron attention oracle is to collect (or create if they do not exist) multiple binary prediction markets about him, such as “Will LeBron James' fan count exceed X million by the end of this month?”, “Will LeBron win a championship in 2026?”, “Will LeBron be named MVP in 2026?”, etc. A complete oracle should include more underlying markets, but this example will focus on these three. The index price will be calculated by weighted aggregation of the prices, liquidity, settlement times, and event significance of each market.

For each prediction market, we need to consider the following four dimensions: price, liquidity, remaining settlement time, and event significance coefficient. To simplify the explanation, we use a basic weight calculation formula: the significance coefficient of each market ranges from 1 to 10 points, and the weight is calculated by combining liquidity and time factors.

Assuming the importance scores of the three markets are 8, 2, and 10, the weights of each market are calculated as follows:

The final attention index is as follows:

Assuming that the settlement periods for the three prediction markets are 180 days, 20 days, and 180 days, with their event importance coefficients being 8, 2, and 10 respectively, the comprehensive calculation is as follows:

Clearly, there are more complex methods for calculating attention indicators, such as using open interest instead of trading volume, considering related events, adjusting market depth, and accounting for non-linear relationships among variables. We have created an interactive website for readers to build custom indices through the real-time Kalshi market.

The main advantage of this prediction market-based oracle construction method is that manipulative behavior will incur actual costs. If traders go long on LeBron's attention and try to raise the index, they need to buy positions in the underlying binary prediction market. Assuming that the underlying market has sufficient liquidity, this means that they must establish positions at prices deemed high by the market.

Another increasingly important advantage as the market expands is that binary prediction markets provide spot hedging channels for market makers. If market makers short the attention index, they can hedge their risk by going long on the underlying prediction market positions that make up that index.

Adjacent has created a political trend index on Kalshi using real-time liquidity markets (such as Democrat vs Republican, New York City mayoral election, etc.). We believe this approach can be extended to track attention on any topic. As prediction markets develop, the range of viable topics will continue to expand.

3**, Design Trade-offs of Attention Oracles**

Our oracle architecture needs to weigh multiple factors. From a more macro perspective on attention oracles, the following are the core consideration dimensions:

• Strength of correlation of input data;
• Practical feasibility of data acquisition;
• The manipulability level of input variables;
• Algorithm function design for calculating attention index.

The most significant trade-off in the oracle solution we propose lies in the difficulty of data acquisition. To build an oracle for LeBron James's attention, multiple high-liquidity prediction markets need to be created for related topics, and these markets must maintain liquidity continuously and be promptly replaced when old topics become invalid. Therefore, this design is only suitable for niche, highly followed topics that already have mature prediction markets (such as Trump or Taylor Swift).

Another contradiction lies in the fact that regardless of the outcome of the event, attention may still increase. For example, even if LeBron fails to win another championship, discussions about his declining status may actually heighten attention. In the real world, attention often flows towards unexpected events, while prediction markets only measure the probability of an event occurring—if the market expects LeBron to win the MVP but he does not, public discussion may become more fervent when the index drops, and fans will argue about the unfairness of the selection.

The optimal solution may be a hybrid approach that combines prediction markets, social media, and other data sources. Google Trends recently opened its search trend API to developers, and the search volume is clearly related to attention, with the deduplication mechanism making it more resistant to manipulation than social media metrics. LLM can also be used to analyze easily manipulable data sources (such as mainstream media headlines or trending posts on the X platform) and filter out junk information, thereby building a more robust assessment system.

We believe that mature exchanges like Kalshi and Polymarket are best positioned to launch attention perpetual contracts, as they already have a significant amount of liquidity in their underlying markets and a large trading user base. However, the opportunities for attention assets are not limited to industry giants.

One feasible solution is to establish a treasury specifically for trading prediction markets, aimed at going long/short on specific themes. For example, the “Long Taylor Swift Treasury” could buy “yes” contracts for events such as her songs entering the top ten, Super Bowl performances, etc., with the treasury manager determining which markets are related to increased attention.

Another mode is to utilize Hyperliquid's builder-deployed perpetual contract feature. The HIP-3 proposal gives market deployers the flexibility to define the oracle—allowing them to construct indices using a combination of Kalshi/Polymarket prices, social media metrics, Google search trends, news headlines, and other data sources.

4**, The Potential of Attention Assets**

Ironically, the first mature application scenario of the attention economy may appear in the stock market. Stock prices consist of two main elements: discounted cash flow value (i.e., intrinsic value) and meme value.

Historically, most stocks did not have significant meme value. However, in recent years, with the rise of forums like WallStreetBets and retail trading platforms like Robinhood that operate 24x5, more and more stocks have begun to consistently carry meme value.

The core task of a stock research analyst is to determine stock prices. While there are established methods for calculating DCF components, how can we quantify meme value? As more assets are traded based on meme value, developing modeling methods for meme value is imperative. Professional investors have begun to use metrics such as fan count, likes, and exposure to assess market sentiment, while predicting the market and establishing connections with other oracle institutions can become an effective tool for measuring stock attention and optimizing trading models.

But the potential of attention assets goes far beyond stock pricing. We believe that predicting attention is an economically valuable activity—attention is a leading indicator of consumer preferences and spending. Companies allocate R&D, hiring, and marketing budgets based on the flow of attention, and the key lies in establishing new heuristic models to track this traffic.