From April 24 to 27, the market seemed to have hit the fast forward button—research reports, financing announcements, token announcements, and macro forecasts exploded onto the screen almost simultaneously, with multiple lines of interest in AI, on-chain assets, and macro bets heating up together, as funds jumped between different tracks, revealing a common emotional tension.

On April 24, DeepSeek released a preview of the V4 series large models, V4-Pro and V4-flash, igniting a new round of "computing power imagination." Almost simultaneously, Lyon interpreted this release in their research report as a smooth advancement of local advanced process capacity expansion in China, continuing to express optimism about SMIC and Hua Hong, while linking Texas Instruments and Intel's first-quarter performance and second-quarter guidance, which were better than expected, with recovering chip demand and rising price trends, attempting to explain the chain of "model upgrade → computing power demand → wafer fab capacity and price increase" to the secondary market. Meanwhile, Four Pillars announced the completion of Series A financing, with Pantera Capital and Further on the investor side, while Guangzhou set up an AI industry investment fund of about 500 million yuan—ranging from international crypto funds to local government funds, bets surrounding computing power, models, and infrastructure emerged concentrated in the same time window.

Another front acted more like a naked token gamble. The Humanity Protocol Foundation, during the same period, requested early investors to adjust the $H token allocation arrangement. This notice brought the project party, early capital, and secondary market expectations to the same table. On April 26, an address associated with Jump Trading transferred a large amount of $H into a centralized exchange, interpreted by the market as a signal of potential selling pressure and changes in large holder strategy amid unclear allocation rules; on the other side, the Curve team attempted to use a special pool of permissionless, redeemable vault tokens to digest about $700,000 in CRV-long bad debt, designing a mechanism to exchange for market funds to take over old debts. How token allocation rewrites the power structure, how large holders take action under uncertain rules, and whether the protocol can rely on institutional design to repair past risks, these questions intertwined within the same few days.

On a higher level, macro narratives and “event betting” were also heating up. A study from London Business School and Yale University was frequently cited in late April: in prediction markets, about 3% of informed traders contributed most of the price discovery, while most average participants consistently lost money, with the losses ultimately flowing to this small group. Spencer Bogart provided figures that concretized this business—a forecast of about $7.5 billion in non-sports category transactions on Polymarket by March 2026, while Kalshi recorded about $1.6 billion, slicing political, macro, and regulatory expectations into contracts that could be bet on. Almost simultaneously, Sam Altman proposed five core principles of AGI from OpenAI, attempting to set a framework of values and governance for AI development, while Citigroup raised its average Brent crude oil price expectation for the second quarter to $110 per barrel, presenting pathways for the third and fourth quarters and the potential of $150 under extreme scenarios, providing new macro anchors for energy and inflation-related assets—this information would ultimately be repeatedly utilized in prediction markets and on-chain derivatives by that 3% of "informed" or self-proclaimed informed traders.

Compressing these clues into the four days from April 24 to 27, the computing power story, local chips, project tokens, and prediction markets appear to be different tracks, yet they carry the same batch of funds and emotions chasing volatility and narratives globally: they set prices for AI models and wafer fabs on one hand, while competing for first-mover advantage in the liquidity pools of $H and CRV on the other, and simultaneously expressing judgments on interest rates, oil prices, and policies in Polymarket and Kalshi. Moving forward, the story will unfold along three conflicting lines: the misalignment between DeepSeek as a catalyst, local chip capacity, and AI model ambitions; interests among large holders, early investors, and secondary participants in the adjustment of $H allocation; and in the rapidly expanding prediction market, only about 3% of informed traders truly dominate price discovery, while the remaining participants merely continue to foot the bill for the game.

DeepSeek ignites, local chips and AI funding heat up together

On April 24, 2026, DeepSeek released the V4 series preview—the V4-Pro and V4-flash. On the surface, this was a technical showcase of "model parameters, inference speed, and benchmarks"; however, Lyon, in a subsequent research report, regarded it as a collective debut from the production side: to deliver such models along this timeline suggests that local advanced process capacity expansion in China is not merely a paper exercise, but is progressing according to plan. In other words, the model release was seen as indirect evidence of wafer fab yield and capacity ramp-up.

Following this judgment, Lyon continued to highlight SMIC and Hua Hong in their research report, believing that these two companies would directly carry the amplified effects of local AI and chip demand. The logic is not complicated: deploying large models requires computing power, the foundation of which lies in chip capacity across different processes and technology nodes, and if local supply can keep pace, orders need not route offshore. More critically, Lyon's perspective transitioned from China to global—within the same report, they noted that Texas Instruments and Intel's first-quarter performance and second-quarter guidance both exceeded expectations, drawing parallels to signs of recovery in chip demand, with some product lines already beginning to show price increases.

This chain was thus connected: from leading models like DeepSeek V4, to the local wafer fabs' advanced process capacity expansion, to the orders and price increase expectations from chip giants like Texas Instruments and Intel—the funds saw a complete path of "model popularity → computing power consumption → upstream chip shipment." Model releases were no longer just a stage for R&D teams but were treated as a signal light for the turning point in the chip cycle.

Almost within the same time window, the primary market and local capital also began to align around the same theme. Four Pillars announced it completed Series A financing, with investors including Pantera Capital and Further; although the specific amount was not disclosed, at this point, it looked more like a declarative chip: international funds were betting on a batch of projects viewed as AI-related infrastructure, hoping to take the critical position in future computing power and data pipelines ahead of time. Meanwhile, Guangzhou established an AI industry investment fund of about 500 million yuan, clearly aimed at supporting local AI-related industries, directly throwing funds at the issue of "who will provide the foundation for this wave of models."

Thus, a clearer picture emerged: at the top, models like DeepSeek V4 keep "breaking out," in the middle, local advanced process capacity at SMIC and Hua Hong is treated as a core chip of a new round of stories, at the bottom, the global chip recovery signal from Texas Instruments and Intel's performance and guidance, and one layer further, the funding entries like Four Pillars' Series A and Guangzhou's 500 million yuan industry fund are opening simultaneously. Each highlight of AI models is not just a contest among engineers, but leads to a layered game among chip capacity, primary equity, local industry funds, and secondary market expectations.

Allocation Rewrite and Large Holder Transfers: The First Pressure Test of $H

As the stories of chips and industry funds began to unfold in the primary market, the Humanity Protocol was already laying out the first round of competition at the level of its tokens. The foundation proposed adjustments to the allocation arrangement of the $H token for early investors; the specific options have not yet been disclosed, but the information itself is sufficiently clear: the economic rights originally written in the investment agreement are being renegotiated. For early capital, this means that lock-up timelines, available chips, and exit paths can be rearranged; for the team, this is an opportunity to strive for more long-term chips and governance discourse after the project narrative amplifies and external expectations rise.

Allocation rewrite is an action filled with tension. From the foundation's perspective, it can be packaged as "more beneficial for long-term health" structural optimization; from early investors' perspective, it can easily be understood as the redistribution of surface rights. $H is in the early stages of competition, and any adjustment regarding allocation is not just a technical term but signals to the market: this set of rules, under pressure, is it elastic, or is it solid? When rules are opened up and rewritten, the community will instinctively ask—next time, will it be their share that's up?

During this window of uncertainty, another force gave its actions on-chain. On April 26, 2026, an address associated with Jump Trading transferred a large amount of $H tokens to one or more centralized exchanges. As a large quantitative and market-making firm, Jump’s reallocation behavior has always been viewed as a signal source of “smart money”: they rarely explain their motivations, but every large transfer tends to be magnified by the market's interpretation. At a time when allocation rules have not yet materialized and the details of terms have not been made public, large holders transferring chips from on-chain to exchanges is naturally interpreted by some participants as a prelude to potential selling pressure.

It is important to emphasize that currently public information has not disclosed the specific data of $H's price trends and on-chain trading volumes; the only hard facts available are: the foundation intends to rewrite early allocations, and large institutions are transferring chips to exchanges. Thus, price expectations are more pieced together from narratives—on one hand, the project party is attempting to extend the time during which “chips are in friendly hands,” while on the other, large holders, regarded as key liquidity providers, are seen pushing chips towards a realm ready for redemption. Under such tension, even without explicit sell-off actions, the shadow of expectations is enough to depress risk appetite.

For the community, this is not only the first round pressure test regarding price but also a test of trust. With allocation terms rewritten, participants will speculate: this time is negotiating with early institutions, but next time will it be aimed at the team, advisors, or even a systemic adjustment regarding future airdrops and ecological incentives? When the funding paths of large holders like Jump overlap with the events of rule adjustments in the timeline, many will instinctively connect the two, constructing a narrative of “internal competition, external takeovers.” Even if this narrative may not be true, it will, in turn, shape liquidity and participant structures in reality.

What ultimately decides whether $H can navigate this round of uncertainty is not a few large transfers on-chain, but the next two matters: first, what does the allocation plan ultimately look like when the foundation and early investors come to terms—is it more long-term chips being ceded to the future of the protocol, or does it maintain a capital-friendly structure; second, with the implementation of the allocation, will there be substantial changes in the on-chain holding structure, is large holder concentration gradually diluted, or is it further amplified? Only when the boundaries of the rules are redefined and the distribution of chips is made visible, $H has the opportunity to emerge from this first round of competition, equipped with an examined pricing logic and a story that can continue to be trusted.

$700,000 Bad Debt Self-Rescue: Curve Bets on Market Takeover

As $H seeks a new equilibrium amid the tug-of-war over “whether the rules should be rewritten,” the established DeFi protocol has already begun to settle accounts for its past leverage frenzy. The roughly $700,000 in CRV-long bad debt on Curve's balance sheet represents a typical history of risk accumulation from leveraged trading: it starts with an amplification of trends—bulls leverage, borrowing tokens to bet on price increases; once the trend reverses, incomplete liquidations and insufficient collateral valuation lead to liabilities creeping up from individual accounts onto the protocol's balance sheet, ultimately resulting in a "nobody's claiming" hole.

The self-rescue path that the Curve team is now presenting is to transform this bad debt from a "shadow asset" into a series of tradable tokens. They proposed establishing a special exchange pool for redeemable vault tokens, packaging the bad debt into the vault, then minting corresponding tokens, making this token a "vehicle" for the bad debt—those willing to take it can exchange real funds in the pool for this risk exposure. The protocol can provide market mechanisms such as discounts in design, allowing early participants to gain potential profits as compensation for taking on the uncertainty.

The key to this design lies in two "don’ts": no permissions, no off-chain agreements. There’s no whitelist, no private negotiations; any address can directly enter this pool on-chain, playing the role of a "taker" or arbitrageur—whether looking at the discount space or the potential upside brought by future "repairs,” pricing is entirely left to on-chain liquidity to derive. Bad debts are no longer abstract figures debated repeatedly in governance forums, but are instead broken down into tradeable risk certificates, displayed on an order book visible to everyone, and everyone can place orders.

On the surface, this is an attempt to make risk extremely transparent and decentralized: the protocol no longer decides who pays, but spreads the opportunities and burdens across the entire market. But from another perspective, it is also a referendum on confidence—how willing is the market to take on this $700,000 CRV-long bad debt at what discount essentially scores the Curve's governance and repayment capabilities: if there is confidence that the team can operate the vault back to break even, the discount will narrow; if confidence falters, these vault tokens will need to find liquidity at even deeper discounts.

More subtly, the proposal currently does not provide a clear implementation timeline; bad debt repairs are still at the proposal stage. The protocol has laid its chips on the table; what remains to be seen is whether on-chain funds are willing to participate as the counterparties in this "market-based restructuring"—and whether the entire DeFi world is still willing to place trust in a mechanism that saves itself using code.

3% Informed and $7.5 Billion Transactions: The True Face of the Prediction Market

If the bad debt restructuring in DeFi is still a game “between codes,” then the prediction market takes all bets back to the most primordial question: who gets the information earlier, who dares to place a hefty bet before everyone else.

The study from London Business School and Yale University breaks down this game's mechanics crisply: in the prediction markets they observed, only about 3% of informed traders contributed the vast majority of price discovery, while the overwhelming majority of average participants consistently stood on the losing side, and these losses almost all flowed to the small group of informed traders. In other words, the operation of “collective wisdom” in the market does not mean everyone is a bit smarter, rather it turns 97% of the people into a paid sample for the 3% to test the world and validate their views.

The disparity between 3% and 97%, after magnifying the transaction scale, reveals its cruelty. The data revealed by Spencer Bogart from March serves as a key anchor in this discussion: for the month, Polymarket alone had a transaction volume of approximately $7.5 billion in non-sports categories, while Kalshi also recorded roughly $1.6 billion in the same period. These two platforms dealt with contracts surrounding macro, political, and economic events—that is, the pricing subjects are no longer niche topics, but major issues like interest rate trajectories, election outcomes, policy implementations, and economic data that directly impact asset prices.

When $7.5 billion in chips are bet on "where macro and socio-economic trends will head," what the 3% informed traders are doing becomes no longer a niche interest in an academic sense, but a genuine issue of profit distribution. Research has already shown us that most people's losses will be taken by that 3%—in platforms like Polymarket and Kalshi, what is taken is not just a few hundred dollars of exploratory betting chips, but the fuzzy views of the retail traders on "the future” itself.

This is also how the narrative of "collective wisdom" is reshaped in the face of data: superficially, everyone is casting their votes on the future with their own understanding, with the price being the weighted average of the majority's opinions; in reality, prices are more about that 3% using real capital and sharper information to "carve" their judgments onto the order book, and then letting 97% gradually verify over time—those who are wrong pay tuition; those who are right get lifted out of the field by the subsequent influx of capital. The prediction markets appear to be democratic oracles, but in essence, they are arenas dominated by a minority of information advantage, while ordinary players merely serve as the continually entering counterparties.

Bringing the perspective back to those few days in late April becomes clearer: DeepSeek ignited the local AI narrative on April 24 with its V4 series preview; Lyon's report emphasized advanced process capacity expansion, chip demand recovery, and price increase expectations; Sam Altman publicly articulated the five core principles of AGI, attempting to frame the values for technological development; Citigroup raised its forecast for the average Brent crude oil price in the second quarter of 2026 to $110 per barrel, with a potential upward trajectory to $150 in extreme scenarios. AI, chips, energy, inflation—these are overarching narratives that define asset pricing, all heating up simultaneously throughout these days.

Against this backdrop of macro and tech themes being vividly lit, one of the most handy tools for informed capital is the prediction market: they need not hunt for counterparts like traditional derivatives or narrate "opinions" to any regulators; they can directly break down their judgments about the Chinese AI path represented by DeepSeek, the chip cycle, and the outlook for oil and inflation into a series of yes/no, range or time contracts, and then stake their positions in platforms like Polymarket and Kalshi ahead of time. The larger the transaction volume, the more the odds swing towards that 3%.

Thus, the meaning of “betting on future trajectories” is subtly rewritten: for the vast majority, prediction markets serve as an entry point to engage in macro and AI topics, a platform to translate emotions and views into numbers; for that 3%, it becomes a nearly tailor-made profit-harvesting machine—leveraging faster access to news and deeper industry understanding, they set prices on the order book for future pathways in advance, waiting for others to imply "endorse" their predictions with actual losses.

On the surface, this is a participatory voting mechanism about the future for everyone; but following the data thread from the London Business School and Yale, you will find it resembles an amplifier designed specifically for a few informed players: the tremendous uncertainties of AI, macro, and political economy are packaged into a series of tradable “questions,” and those who can continuously benefit from it remain just that 3%.

From AGI Principles to Oil Price Projections: What Kind of Future Will Funds Bet On

Bringing the timeline back to that moment on April 24: DeepSeek launched the V4 series preview—the local large model narrative reignited; Lyon immediately connected this launch directly to the “smooth progress” of China’s advanced process capacity, continuing to name SMIC and Hua Hong, and utilizing the better-than-expected performances and guidance of Texas Instruments and Intel in the first quarter to strengthen the chain of “AI demand → chip shipments → price increase expectations.” Simultaneously, from Four Pillars’ Series A funded by Pantera Capital and Further, to the Guangzhou 500 million yuan scale industry fund, the story of local AI capacity also found quantifiable entry points in the primary market and local funding.

At the same time, the Humanity Protocol Foundation requested early investors to adjust the allocation arrangement of $H, while an address associated with Jump Trading added a large amount of $H into exchanges on April 26, condensing the competition between the project party and early capital into a string of allocation parameters and a record of potential selling pressure on-chain. Looking over to DeFi, Curve proposed using a special exchange pool of redeemable vault tokens to self-rescue that roughly $700,000 CRV-long bad debt—without relying on off-chain negotiations, no permissions needed, directly handing risk and discounts over to liquidity for pricing. In terms of computing power, it’s DeepSeek and local wafer fabs; asset-wise, it’s $H and CRV-long; and information-wise, it’s the batch of participants placing orders with real cash on Polymarket and Kalshi: the same category of funds, betting simultaneously on the same question along the hardware, asset, and information lines—what will the future ultimately look like?

Above all these bets, Sam Altman project another completely different language. In late April, he publicly proposed the five core principles of AGI from OpenAI: democratization, empowerment, universal prosperity, resilience, and adaptability—the tech giant aims to first delineate value boundaries, then accelerate within those boundaries. It addresses normative questions of “what AI should become,” while those on-chain are addressing a colder, harder question: can these principles truly be implemented, when can they be implemented, and if they are implemented, will they be violated? The allocation parameters of $H, the bad debt discount of Curve, the pace of local chip capacity expansion, and even every bet regarding AI, regulation, and geopolitics in the prediction market, essentially provide discounts to these five principles: how much discount is the probability and risk you are willing to take on.

Macro variables amplify all this further. Around April 27, Citigroup raised its forecast for the average Brent crude oil price in the second quarter of 2026 to $110 per barrel; in the same forecast, it noted $95 for the third quarter, and $80 for the fourth quarter; under specific scenarios of continued disruptions in fund flow, a pathway to $150 for oil prices was also written into the report. For those engaged in AI, chips, tokens, or prediction markets, this is not merely an isolated commodity price curve, but rather a core parameter that will seep into valuation models, financing costs, and liquidity discounts: whether oil prices settle at $80, $110, or touch $150 in an instant, each point corresponds to different cash flow assumptions, different risk appetites, hence associated with different expectations for DeepSeek, market capitalization imaginings, $H selling pressure and unlock pace, as well as odds in prediction markets.

If you piece together the fragments from these days, you will find that the 3% of informed traders did not just remain in one market: they reference Lyon and Citigroup reports to judge the extent to which local AI capacity can be realized and how the energy curve may reshape costs and interest rate expectations; they monitor what allocation scheme Humanity Protocol will ultimately present for $H, how much discount the $700,000 bad debt of Curve can be digested by the market; meanwhile, they observe how regulatory and liquidity environments may reshape the scale and participation thresholds of such prediction markets as Polymarket and Kalshi. In the coming quarters, the key intersections that truly need ongoing tracking are these four main lines: whether local AI capacity will be successfully implemented as reported, how the allocation competition of $H will ultimately divide the chips, whether Curve's self-rescue mechanism can clear risk without requiring permissions, and whether prediction markets will evolve into a larger-scale "pricing machine" in the regulatory gaps and amid capital influx, or be forced to shrink into a fringe betting table.

DeepSeek's V4 preview, the expansion of local chips, the restructuring of chips between $H and CRV-long, Citigroup's oil price curve and Sam Altman's AGI principles, are merely different scenes from the same script: one side attempts to outline gentle and idealized value boundaries for the future, while the other is directly interrogating, through capital flow, a more straightforward question—when these boundaries encounter practical challenges such as computing power bottlenecks, asset gambits, and oil price shocks, which future will ultimately be the one priced in.

Join our community for discussions and to grow stronger together!
Official Telegram community: https://t.me/aicoincn
AiCoin Chinese Twitter: https://x.com/AiCoinzh
OKX Benefit Group: https://aicoin.com/link/chat?cid=l61eM4owQ
Binance Benefit Group: https://aicoin.com/link/chat?cid=ynr7d1P6Z

免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。