InfoFi Exposed: Why Tokenizing Attention Is Crypto's Most Controversial Experiment
InfoFi promises to turn information and attention into tradable assets. But ZachXBT calls it 'the most widely promoted scam this cycle.' Here's the full picture on what's working, what's broken, and how smart traders should approach it.
The crypto industry loves inventing new words for old problems. DeFi gave us programmable money. SocialFi tried to monetize followers. Now InfoFi wants to tokenize something even more abstract: your attention.
The concept sounds elegant. Vitalik Buterin introduced "Information Finance" in November 2024, describing it as a discipline where you "start from a fact that you want to know, and then deliberately design a market to optimally elicit that information from market participants." Prediction markets like Polymarket proved this works. The platform processed over $3.3 billion in bets on the 2024 US presidential election and delivered more accurate forecasts than traditional polling.
But something went wrong when the crypto industry tried to apply this concept beyond prediction markets. The "yap-to-earn" platforms that emerged turned Crypto Twitter into a wasteland of AI-generated content, engagement farming, and manufactured hype. ZachXBT, the blockchain investigator, put it bluntly: InfoFi platforms are "the most widely promoted scams in this cycle."
So which is it? A revolutionary new asset class or an elaborate extraction mechanism? The answer is both, and understanding the difference could save you significant money.
Vitalik's original InfoFi thesis was narrow and specific. He proposed that Ethereum could host markets designed to extract truthful information from participants. The mechanism is simple: when people bet real money on outcomes, they have financial incentives to be honest. Wrong predictions cost money. Correct predictions profit.
Polymarket demonstrated this works at scale. During the 2024 election, when traditional polls showed a toss-up, Polymarket consistently priced Trump's victory probability above 60%. Bettors with real money on the line processed information differently than pollsters conducting surveys. The market aggregated thousands of individual assessments into a single probability, updated in real-time as news broke.
The crypto industry saw this success and asked a natural question: if prediction markets can extract truthful information about elections, can we build markets that extract other valuable information? What if we could reward people for quality research, accurate analysis, or insightful commentary?
This is where the theory diverged from practice.
The platforms that emerged, Kaito, Cookie3, Galxe's Starboard, and others, created point systems that attempted to measure "content quality" and "attention value." Users earn points (Yaps, Snaps, or similar) for posting, commenting, and engaging with crypto content. The top performers receive token rewards. In theory, this should surface valuable insights. In practice, it created perverse incentives.
The problem is measurement. Prediction markets have a clean feedback loop. The election happens. Trump wins or loses. Bets resolve. Everyone knows who was right. But what makes a tweet "valuable"? The platforms tried using engagement metrics, AI-scored "quality," and community voting. Each approach proved gameable.
Louround, co-founder at Redacted Research, documented the result: "We've seen through the LOUD experimentation that mindshare does not equal protocol interest, nor value creation." The LOUD project achieved 60% mindshare on Kaito's leaderboards and reached a $30 million fully diluted valuation. Within two weeks, it collapsed to $1.4 million. The attention was real. The value was not.
Understanding how InfoFi gets gamed helps you recognize what's happening in real-time. The farming operations follow predictable patterns.
The most basic operation is volume posting. An account posts dozens of replies daily, tagging projects and using trending hashtags. The content is generic ("Great progress team! 🔥" or "This is bullish for the ecosystem") but algorithms initially counted it as engagement. When platforms cracked down on low-quality replies, farmers evolved.
The next generation used AI-generated content. GPT-4 can produce coherent analysis that passes surface-level quality checks. An operator feeds it a project's documentation, asks for a thread, and posts the output. The content reads well but contains no original insight. It's a sophisticated form of copy-paste that inflates perceived activity around projects.
Coordinated groups take this further. A project pays a network of accounts to simultaneously post positive content. Kaito's mindshare metrics spike. VCs and exchange listing teams see the "organic interest" and take meetings. The project gets funded or listed based partly on manufactured metrics. After the token launch, the coordinated group dumps their allocations. The community that supposedly loved the project disappears.
ZachXBT launched a $5,000 bounty to scrape user data from Kaito Yaps, Wallchain, Galxe, Layer3, Cookie, and Xeet. His goal: expose the coordinated farming operations polluting crypto social media. The fact that a prominent investigator considers this worth paying for tells you how severe the problem has become.
The feedback loop Louround described is insidious. Projects need visible "traction" to raise funds and get listed. Platforms provide metrics that projects can game. VCs and exchanges use these metrics as one input in their decisions. No party has strong incentives to expose the system's flaws because everyone benefits from the appearance of activity.
Zero Knowledge, another Redacted Research member, quantified one symptom: "A guy who drops 900+ replies in a day is not an advocate for your tech or brand. It's an extractor that wants to dump tokens on day one." The platforms reward activity, not insight. The natural result is maximum activity with minimum substance.
Despite the criticism, some InfoFi applications deliver genuine value. The distinction matters because throwing out the entire category means missing real opportunities.
Prediction markets remain the clearest success. Polymarket's $8 billion valuation (following a $2 billion investment from Intercontinental Exchange in October 2025) reflects institutional confidence in the model. The mechanism works because predictions resolve to observable facts. There's no ambiguity about who was right.
The expansion of prediction markets into new domains shows continuing innovation. Sports betting was an obvious extension, but projects now offer markets on crypto prices, protocol metrics, and governance outcomes. Each market creates information that didn't exist before. When thousands of people with money at stake estimate ETH's price one month from now, the resulting probability distribution contains real signal.
Reputation systems represent a second promising category, though still early. Protocols like GiveRep and Ethos attempt to track on-chain reputation, who delivered accurate analysis, who built useful tools, who contributed meaningfully to DAOs. If these systems mature, they could solve a real problem: distinguishing genuine contributors from farmers in airdrop allocations, governance votes, and community grants.
The challenge is that reputation is inherently harder to measure than prediction accuracy. A prediction either happens or it doesn't. Reputation involves subjective assessments that different observers might weigh differently. The platforms attempting this need years of iteration before we'll know if they work.
Data and analytics layers form a third category. Cookie3 aggregates AI agent data across Web3. Kaito Pro (distinct from the controversial Yaps system) provides research tools for professionals. Santiment tracks on-chain and social metrics. These tools sell information products to paying customers, a straightforward business model that doesn't depend on token incentives working correctly.
Looking at InfoFi in historical context clarifies what's actually new and what's recycled from previous cycles.
DeFi (2020) tokenized money. You could lend, borrow, trade, and earn yield using smart contracts instead of banks. The key metric was TVL (Total Value Locked). The innovation was real: billions of dollars now flow through protocols that didn't exist five years ago. Some DeFi products (Uniswap, Aave, Compound) became infrastructure that survived multiple bear markets.
SocialFi (2023) tried to tokenize social connections. Friend.tech let you buy "keys" representing access to creators. Farcaster built a decentralized social protocol. The key metric was engagement and key holder counts. The results were mixed. Friend.tech generated massive initial volume then faded as the novelty wore off. Farcaster raised $150 million and continues building, but mainstream adoption remains limited.
InfoFi (2024-25) attempts to tokenize attention, reputation, and predictions. The key metrics vary: mindshare for Kaito, Yaps for content creators, betting volume for prediction markets. Like SocialFi, the results are mixed. Polymarket is a legitimate success. Yap-to-earn platforms are largely extractive. The category is too broad to judge uniformly.
The pattern across cycles is that each "Fi" starts with genuine innovation, gets over-hyped, attracts extractive actors, crashes, and then the legitimate use cases survive and mature. We're currently in the over-hyped and extractive phase for much of InfoFi. The question for traders is identifying which applications will survive the inevitable shakeout.
If you're trading crypto in 2025, you can't ignore InfoFi entirely. The platforms influence token prices, airdrop allocations, and market sentiment. But engaging with them requires a different strategy than the farmers use.
The first principle is avoiding the farming trap. Spending hours daily posting for Yaps or Snaps puts you in direct competition with coordinated groups running AI-generated content at scale. Even if you're legitimately insightful, your individual posts compete against industrial operations. The time investment rarely justifies the token rewards, which typically work out to below minimum wage when farming becomes saturated.
The second principle is using InfoFi signals selectively. Kaito's mindshare data, despite being gameable, sometimes contains real information. When a project's mindshare spikes, it might indicate coordinated farming, or it might indicate genuine interest. The trick is cross-referencing with on-chain data. If mindshare increases but wallet activity stays flat, the signal is noise. If mindshare increases alongside smart money wallet accumulation, something real might be happening.
The third principle is prioritizing on-chain signals over social signals. Blockchain data is harder to fake than Twitter engagement. When wallets with strong historical performance start accumulating a token, that's meaningful information. When a project's liquidity depth increases, that's meaningful. These signals existed before InfoFi and remain more reliable than gameable social metrics.
This is where tools like EKX.AI's Pre-Pump Scanner fit into a trading workflow. Rather than tracking which tokens are trending on Kaito (which could reflect farming), the scanner monitors actual on-chain activity. Unusual wallet transactions, liquidity changes, and smart money movements often precede price action by hours or days. The signal comes from what money is doing, not what content farms are posting.
InfoFi's promise was surfacing valuable information through market mechanisms. The best applications still do this, but through on-chain data rather than social engagement. Prediction markets work because predictions resolve to facts. On-chain analytics work because transactions are immutable. Social engagement metrics work poorly because they're easily gamed.
Vitalik's original vision for InfoFi remains compelling. Markets that aggregate information and reward accuracy could improve decision-making across many domains. The current implementation problems don't invalidate the concept.
The platforms are iterating. Kaito recently updated its algorithm to "prioritize quality over quantity," excluding posts that only mention rewards or rankings, limiting weekly mindshare tweets, and enhancing loyalty rewards for consistent contributors. Whether these changes work remains to be seen. The adversarial dynamic between platforms and farmers means any improvement triggers an adaptive response from the other side.
Longer term, several developments could make InfoFi more robust. AI-driven content filtering is improving. If platforms can reliably detect AI-generated posts, they can exclude them from rewards. Decentralized identity systems could link on-chain reputation across platforms, making it harder to spin up fresh accounts for farming. Cross-platform reputation scores could create consequences for gaming behavior that persist beyond any single protocol.
The prediction market category will likely continue expanding. Polymarket's success attracted institutional interest and regulatory clarity (in some jurisdictions). More event types will become tradeable. Integration with AI could enable micro-markets on questions too small for traditional prediction markets. These developments align with Vitalik's original thesis.
For the yap-to-earn category, the path forward is less clear. The fundamental problem is that content quality is subjective and engagement metrics are gameable. Until platforms solve this measurement problem, the farming incentives will persist. Some platforms may pivot to different models. Others will fade as users recognize the extractive dynamics.
InfoFi is not one thing. It's a broad category containing genuinely useful applications and obvious scams. Your strategy should differentiate between them.
Prediction markets (Polymarket, Kalshi, Azuro) have proven value. If you're interested in this subcategory, use them for what they're good at: aggregating information about discrete, resolvable events. The odds on Polymarket often contain more signal than news articles or Twitter threads.
Yap-to-earn platforms are generally not worth your time as a participant. The farming economics favor industrial operations over individual contributors. If you use these platforms at all, use them as one input among many for gauging market sentiment, not as a reliable source of alpha.
On-chain analytics tools remain underrated. The information advantage comes from seeing what money does before the crowd notices, not from seeing what content farms post. Prioritize tools that surface wallet activity, liquidity changes, and transaction patterns.
The attention economy is real. Crypto projects live and die by their ability to capture attention. This creates trading opportunities when you can identify genuine attention spikes before they're reflected in price. The challenge is distinguishing genuine interest from manufactured hype. On-chain data helps. Social metrics, used carefully, can supplement. But don't mistake high mindshare for guaranteed price appreciation, as LOUD's collapse demonstrated.
The InfoFi narrative will evolve. A year from now, the terminology might change, specific platforms will rise and fall, and new applications will emerge. The underlying dynamics, the value of information, the incentives to game metrics, and the difficulty of measuring quality, will persist. Understanding these dynamics matters more than mastering any specific platform.
Attention is valuable. Information is valuable. But the systems designed to capture and trade that value remain deeply flawed. Navigate accordingly.
