InfoFi's Reckoning: How One API Ban Reshaped Crypto's Trillion-Dollar Bet on Information
On January 9, 2026, bots posted 7.75 million crypto-related messages on X in twenty-four hours — a 1,224% spike above baseline. Six days later, X's product lead Nikita Bier walked to a microphone and ended an entire crypto sub-sector with one announcement: the platform would permanently revoke API access for any application that financially rewards users for posting. Within hours, KAITO and COOKIE — the two flagship tokens of the so-called Information Finance movement — fell more than 20%. The sector that bullish analysts had spent twelve months calling "crypto's next trillion-dollar category" suddenly looked like a permissioned business with a single landlord.
Three months later, the obituary writers look premature. Polymarket and Kalshi are clearing roughly $25 billion in combined monthly volume. Grass, the bandwidth-sharing data network, has crossed three million active nodes scraping the open web for AI training corpora. And Kaito itself, after sunsetting its incentivized "Yapper Leaderboards" in January, came back in February with a Polymarket partnership that turned attention itself into a tradeable derivative. InfoFi did not die. It molted — and the version that survived looks structurally different, and structurally healthier, than the one investors were pricing at peak hype.
What InfoFi Actually Is — and Why It Looks Like Three Different Industries
The slogan that defines the movement is short: information should pay rent to the people who produce it. The implementation is fragmented across three sub-sectors that share an economic logic but almost no technical surface area.
Attention markets measure who is shaping conversation and pay them in tokens for that influence. Kaito, founded in 2022 by former Citadel hedge-fund manager Yu Hu and backed by Dragonfly Capital, Sequoia Capital China, Jane Street, and HashKey Capital, anchors this category. Cookie.fun ranks AI-agent mindshare. Bankr and a long tail of "post-to-earn" platforms compete for the same wallet of crypto-native creators. The KAITO token, which crossed a $1.9 billion fully diluted valuation at airdrop, today trades around $452 million FDV — a roughly 75% drawdown that captures the sub-sector's vulnerability.
Data-labor protocols invert the Web2 bargain by paying users for the raw material that trains AI. Grass, the largest by user count, lets households share idle internet bandwidth in exchange for tokens; the network grew fifteen-fold in 2024, from 200,000 to 3 million users across 190 countries, and now operates 100 Gbps of aggregate bandwidth with a roadmap toward one petabyte per day. Vana, an MIT spinout that launched its mainnet in December and now hosts more than 12 million data points across multiple "Data DAOs," introduced a VRC-20 token standard so individual datasets can themselves trade as on-chain assets. Sahara AI rounds out the trio with a node-based incentive program for AI-ready data labeling.
Prediction markets are the oldest of the three and, in 2026, by far the largest. Polymarket and Kalshi together generated more than $44 billion in 2025 trading volume, then accelerated: January 2026 set an all-time peak of $26.75 billion in monthly notional, and March closed at $25.7 billion — almost thirteen times the $2 billion booked in March 2025. Through April 20, Kalshi's year-to-date volume sits near $37.49 billion against Polymarket's $29.23 billion, with Polymarket reportedly raising at a $15 billion valuation despite trailing its rival in sustained turnover.
The thesis that ties these three together is simple: in every case, a substrate that Web2 platforms gave away for free — engagement, data, opinion — is being repriced as a yield-bearing asset, with the cash flow accruing to producers rather than aggregators.
The Attention Economy's Near-Death Experience
To understand why January 15 was an extinction-level event for a meaningful slice of InfoFi, you have to understand how thin the technical foundation was.
Every "post-to-earn" application in the Kaito orbit relied on the same plumbing: scrape X via the developer API, run an internal model that scored each post's reach and engagement, mint tokens proportional to the score. The model was the IP. The scraping was the revenue. And both depended entirely on a third party — X — choosing not to turn off the spigot.
When the spigot turned off, the fragility was total. Kaito sunset YAPS and the Yapper Leaderboards on January 15, the same day Bier's announcement landed. Cookie DAO ended its Snaps product. A long tail of imitator projects — Xeet, BubbledMaps, Loud, Arbus — saw their tokenomics simply stop working. CryptoQuant CEO Ki Young Ju captured the rationale in two numbers: 7.75 million bot-generated crypto posts in a single day, a more than twelve-fold increase over baseline. The InfoFi reward loop had created the precise externality — synthetic engagement designed to game the leaderboards — that made hosting it untenable for the underlying platform.
Kaito's response, announced in February, was a strategic pivot rather than a retreat. The new product stack is four pieces: Kaito Pro (institutional research terminal), Kaito Studio (creator analytics without the financial reward layer), Capital Launchpad (a token-distribution venue), and Kaito Markets (an attention-derivatives venue built on Polymarket). The Polymarket integration is the conceptually interesting move: instead of paying users to generate attention, Kaito now lets traders wager on which brands, narratives, or public figures will dominate the conversation next. Attention itself becomes the underlying asset; the platform becomes an exchange rather than a payroll.
This is not a small philosophical shift. Wagering on attention does not require X's API to function — derivative pricing is endogenous to the market itself. The post-API-ban version of InfoFi, in other words, looks much more like a financial-markets business and much less like a media-incentive program. That is probably the only configuration that survives a Twitter willing to pull the rug whenever bot density gets uncomfortable.
Data Labor: The Quietest Pillar Is Also the Most Defensible
While the attention sub-sector absorbed the headlines, the data-labor protocols quietly compounded. Grass alone records weekly scraping volumes measured in the thousands of terabytes — as of early March, a seven-day cumulative throughput of 6,694 TB — and is positioning Phase 2 of its rollout as the first crypto network capable of sustaining multi-petabyte daily workloads, including video data for vision-model training.
The structural difference matters. Where attention markets are a layer of metadata sitting on top of someone else's platform, data-labor networks own their physical substrate: residential bandwidth, idle compute, user-uploaded datasets. There is no API that X (or anyone) can revoke to make Grass stop functioning. A government could in principle shut the network down at the legal level, but doing so would require an enforcement model the size of the residential-VPN industry — an order of magnitude harder than telling a single platform's developer-relations team to flip a flag.
Vana extends the same logic into a different domain. Its DataDAOs let users contribute personal data — health records, social-graph exports, browsing histories — to topic-specific pools that license the aggregated corpus to AI labs. The VRC-20 standard, introduced in April 2025, means each DataDAO can issue a token that represents a proportional claim on its dataset's revenue stream, with fixed supply, on-chain governance, and explicit liquidity rules. This is closer to a securitization framework than a token launch: the data has a measurable yield, the yield has a tradable wrapper, and the wrapper has a regulatory shape that institutional buyers can underwrite.
The MIT-incubated provenance also gives Vana an unusual asset: an academic research narrative that translates well to enterprise procurement. More than one million people have contributed data to Vana's network. Sahara AI is pursuing a parallel strategy with its Knowledge Base, building infrastructure for the data-labeling tasks that frontier models still need humans to perform.
Prediction Markets: The Truth Machine Goes Mainstream
Of all three pillars, prediction markets are the one that even crypto skeptics now acknowledge has crossed the chasm. The volume numbers are the easiest tell. Eight consecutive weeks of Polymarket clearing more than $2 billion. A combined $25 billion monthly average across the top two platforms. A volume curve that grew thirteen-fold year-over-year in March.
The structural question for the next twelve months is whether prediction markets are best understood as a gambling vertical, a financial-data vertical, or a media vertical — because each framing implies a wildly different valuation multiple.
The gambling framing is what most regulators and many investors default to. By that logic, Polymarket and Kalshi are best benchmarked against DraftKings, FanDuel, and the global sports-book industry. The prize is large but the multiples are modest, and the regulatory overhang is permanent.
The financial-data framing is the one Wall Street is increasingly testing. In this view, prediction markets are the fastest-pricing source of probability on every major future event — elections, Fed decisions, geopolitical conflict, sports outcomes — and the natural buyers are hedge funds, news desks, and forecasting platforms that need real-time probability feeds the way they currently need Bloomberg terminals. Polymarket's reported $15 billion valuation only makes sense under this framing.
The media framing is the most speculative and possibly the most consequential. If prediction markets become the "truth machine" that arbitrages every contestable claim — a function some commentators are now arguing is the actual long-term TAM — then the comparable is not DraftKings but rather the entirety of opinion journalism. That is what the more aggressive 2026 analyses mean when they project a $10 trillion annual notional volume target: not gambling on Sunday football, but a structural replacement for the institutions that currently certify what is true.
The discount that public-market investors are reportedly applying to Polymarket relative to Kalshi — driven in part by Polymarket's deeper crypto associations — suggests that the financial-data framing is winning at the margin, but the multiples have not yet committed to the media framing. That gap is the alpha for whichever investor is right about which framing prevails.
The Risk Stack Is Real, and Mostly Concentrated in One Sub-Sector
The InfoFi bull case has three structural risks, and they are not evenly distributed.
Sybil resistance is the deepest issue for attention markets. The January reset only happened because the leaderboards were, in fact, gameable to the point of breaking the host platform. Every redesign of YAPS, Snaps, and their successors will face the same problem: any economically meaningful reward for "engagement" attracts adversarial automation faster than any classifier can keep up. The Polymarket pivot sidesteps this by making attention the underlying of a derivative rather than the trigger for a payout, which is structurally better — but it does not solve the sybil problem so much as route around it.
Platform dependency has been substantially demonstrated. The lesson of January 15 is that any InfoFi project whose business depends on a single Web2 API surface is, in a meaningful sense, a tenant rather than an owner. Data-labor protocols mitigate this by holding their own substrate. Prediction markets mitigate it by sourcing their inputs from the underlying world rather than from any platform. Attention markets remain exposed unless they can decouple measurement from any single source — and so far the Kaito playbook for doing so is unproven at scale.
Regulatory shape is the long-tail risk. Prediction markets sit at the intersection of CFTC jurisdiction, state-level gambling law, and ongoing federal-court litigation over event contracts. Data-labor protocols face GDPR-style data-licensing regimes in every major jurisdiction. Attention markets face securities scrutiny if their tokens look too much like investment contracts. None of these risks is acute enough to break the sub-sector, but each is large enough to compress multiples over a multi-year horizon.
What the April 2026 State of Play Tells Us
Three months after the crash, the picture that emerges is a barbell. On one end, prediction markets have crossed into mainstream financial infrastructure, with volume curves and institutional engagement that no longer require apologetic framing. On the other end, data-labor protocols are quietly becoming part of the AI supply chain, with Grass and Vana operating at scales — three million nodes, twelve million data points — that put them within striking distance of being load-bearing infrastructure for the next generation of foundation models.
In the middle sits the attention sub-sector, which had to take a brutal lesson in January about the difference between a business model and a metadata layer, and which is now rebuilding around financial primitives rather than reward primitives. Whether Kaito Markets and its imitators succeed in turning attention into a functioning derivatives venue is still an open question. Whether attention itself will remain a tokenizable asset — in some form, on some platform — is much closer to settled.
The 2026 InfoFi story is not the trillion-dollar parabola that the most enthusiastic analysts pitched in late 2025. It is something more interesting: a category that was forced through its own DeFi-Summer-style purge before it had time to fully inflate, and that emerged with a cleaner accounting of which sub-sectors actually have economic gravity and which were riding the cycle's narrative momentum. The survivors look like real businesses. That is not the most exciting outcome an early investor can imagine, but it is the one that compounds.
BlockEden.xyz operates enterprise-grade RPC infrastructure across the chains that InfoFi protocols depend on, including Solana, Base, and the EVM ecosystems where Kaito, Polymarket, and Vana settle their on-chain activity. Explore our API marketplace to build the next generation of attention, data, and prediction markets on infrastructure designed for production load.