A strange event passed through my terminal last week: ZhiPu, one of China's premier large language model companies, saw its token (or stock — the line blurs in blockchain-native media) surge 19% after a major lockup expiration. Wall Street analysts cheered. The headline wrote itself: 'Unlock fear debunked as capital faith holds.' But anyone who's spent years tracing wallet clusters knows: price is not proof.
Hook
The standard lockup narrative is simple: insiders, employees, early VCs get to sell — they sell, price tanks. That's the signal. But ZhiPu flipped it. Price up 19%. Volume up 300% in the first 24 hours of unlock. Blockchain media outlets latched onto it, framing the move as a vote of confidence from 'smart money' and Wall Street heavyweights. It smelled like a narrative change, not a data fact. So I ran the query.
Context
ZhiPu, for the uninitiated, is the company behind GLM-series models — arguably China's closest equivalent to OpenAI in terms of research pedigree and open-source strategy. Its token (let's call it ZHI) has been trading on a few decentralized exchanges and centralized platforms that accept crypto deposits. The lockup event released approximately 25% of the circulating supply, coming from early backers and team wallets. Typical unlock events in this space — even for well-regarded projects — trigger 30-60% drawdowns within two weeks. The fact that ZHI did the opposite is not just interesting; it's an anomaly worth dissecting.
The blockchain media coverage was glowing: 'Wall Street continues to back ZhiPu,' 'Unlock selling absorbed by strong demand.' But where did the demand come from? And was it real? To answer that, I needed to go beyond the headline and into the chain itself.
Core Insight: The Evidence Chain
I started by pulling transaction data from the block explorer for ZHI's contract. The first thing that jumped out: the volume spike wasn't distributed across thousands of addresses. It was concentrated. In the first 6 hours after unlock, three wallet clusters (groups of addresses linked by common funding sources) accounted for 74% of all buy-side volume. That's not retail enthusiasm. That's accumulation by a small number of actors.
I traced those clusters back using my old method from the 2017 ICO audit days — manually following ETH flows from exchange hot wallets. Two of the clusters sourced their funds from a single Binance withdrawal address at 2:13 AM UTC, just minutes before the first large buy order. The third cluster was even more suspicious: it cycled funds through a Tornado Cash-like mixer before hitting the ZHI/USDT pair on Uniswap V3. The behavior pattern screamed orchestration, not organic demand.
Next, I looked at the selling side. Were insiders actually selling? If so, the price should have dropped unless the buys overwhelmed. I identified 12 labeled team and early-investor wallets. Four of them moved tokens within the first hour of unlock. Combined, they sold 1.2 million ZHI — approximately $4.8 million at pre-unlock price. Interestingly, two of those sales went to the same exchange address that later funded the buy clusters. That's a classic wash trading pattern: sell to yourself, create a price floor, let the market see stability, then attract real buyers.
But the Wall Street piece? I called a contact at a major sell-side firm who covers Chinese tech. He told me no official report had been released on ZhiPu in the past month. The 'Wall Street continues to back' claim was likely a repackaged quote from a note three months ago. The media was attaching a stale institutional endorsement to a live market event. This is common: blockchain outlets know that institutional name-dropping drives engagement, even if the quote is out of context.
I then compared the on-chain data to the price chart. The 19% run happened in three sharp legs, each coinciding with a large buy order from the clustered wallets. Between legs, the price drifted sideways on low volume — no natural buying pressure. This is the signature of a scripted pump, not fundamental revaluation.
To verify, I checked the transaction Liveness index — a metric I built during the DeFi Summer of 2020 to distinguish bot-driven volume from human activity. The index looks at transaction inter-arrival times: bots produce near-uniform intervals, humans produce bursty, clustered timestamps. The buy-side transactions on ZhiPu’s pair showed a standard deviation of less than 0.4 seconds across 150 trades. That's not human. That's an algorithm.
Let’s be clear: none of this proves the company itself is fraudulent. ZhiPu remains a technically strong AI firm. The anomaly is in the market structure around its token. What we're seeing is a setup — likely by a small group of market makers or speculators — to engineer a narrative of strength to attract retail and perhaps subsequent institutional interest. The blockchain media, hungry for a positive story in a bear market, amplified it.
Contrarian Angle: Correlation ≠ Causation
The obvious counterpoint: maybe the buying was genuine institutional conviction. After all, AI is hot, ZhiPu is top-tier, and a 19% pop in a thin market isn't impossible without manipulation. Maybe the three clusters were three different family offices that all independently decided to buy within the same hour using the same exchange source. Improbable but possible.
But the on-chain data doesn't support that. When genuine, unsolicited institutional buying happens, it typically exhibits two signatures: over-the-counter (OTC) settlement that doesn't hit spot markets immediately, and gradual accumulation over days or weeks, not minutes. The ZhiPu buy orders were executed on public pairs in consecutive blocks. No OTC, no gradual positioning. If a real institution wanted a 19% return, they'd buy over a month, not an hour. The compression of buying into a 6-hour window is a hallmark of a coordinated campaign, not due diligence.
Furthermore, the Wall Street 'endorsement' was actually a quote from a strategist who covers Chinese equities broadly, not specific to ZhiPu. The quote itself was three months old, originally given after the company's previous funding round. The media recontextualized it as a current opinion. That's not a lie, but it is deceptive framing. As I always say: trust the hash, not the headline. In this case, the hash (transaction hashes) tells a story of three wallet clusters orchestrating a tight window of artificial demand.
Another blind spot: the lockup itself. Did the unlock really increase circulating supply by 25%? In crypto, 'unlock' can be gamed — tokens might be unlocked but never moved, or moved to custodial wallets that don't sell. I checked the supply schedule. The 'unlocked' tokens were indeed released from the vesting contract, but only 30% of that unlocked supply was transferred out of the original wallets in the first week. The rest stayed put. So the actual sell pressure was only ~7.5% of total supply, not 25%. The narrative exaggerated the 'test' by a factor of three. That makes the 19% pump less impressive — absorbing a 7.5% sell-off is common in liquid markets.
The contrarian take: the 'unlock pump' is a manufactured signal, designed to reset market psychology. The real question is whether the orchestrators intend to hold for long or dump on the narrative peak. Given the short-term behavior pattern (rapid accumulation via bots), my bet is on a staged exit in the next 2-4 weeks. The on-chain evidence chain: bot-like buy pattern + media amplification + stale institutional quote = classic pump-and-dump skeleton.
Takeaway: What to Watch Next Week
The most telling metric isn't price — it's the wallet retention. If the three clusters start distributing their ZHI holdings to new, unlinked addresses over the next 10-14 days, the orchestrated pump is transitioning to a distribution phase. If instead they park the tokens in cold wallets and liquidity deepens with new independent buyers, maybe I'm wrong.
But I've seen this pattern before. In 2021, I uncovered an NFT wash trading ring that used 200 wallets to pump a blue-chip collection by 40% before the floor collapsed. The on-chain signature was identical: concentrated buying timed with media releases, then a slow bleed. ZhiPu's token shows the same early signature.
Set a watchlist alert on the top 5 accumulation addresses. If they go dormant or start sending to exchanges, the party is over. If the narrative shifts to 'ZhiPu token is the next AI powerhouse,' ask for the on-chain receipts. Chaos is just data waiting for the right query. This time, the query returned a red flag.
Yields don't lie. The bots bought. The media sold. Now we wait to see who got the better deal—but I'm not betting on the headline.