I still remember the week that number started circulating in group chats. Nearly $300 million, parked across a cluster of wallets that blockchain researchers said belonged to one venture shop and the people around a token that had lived and died in a single afternoon. Fast forward to early October 2026, and the same cluster is being quoted at roughly $2 million. That is not a rounding error. It is the kind of gap that makes ordinary holders ask a blunt question: where did the rest go?
The honest answer is messier than a headline. A wallet balance is not a bank statement. Token prices can evaporate without a single coin leaving an address. Liquidity can be pulled, swapped, bridged, or frozen. Labels can shift as analysts add or drop addresses. And court files, which move slower than memes, have already closed one civil path in the United States while a separate criminal inquiry continues elsewhere. If you only watch the big number, you will miss the plot.
The $300 Million Figure Was Never Just Cash
Back in mid-February 2025, a blockchain analytics firm said it had mapped more than 1,000 addresses tied to Kelsier Ventures and Hayden Davis. At the moment that map went public, those addresses were valued at close to $300 million. Most of that mark came from LIBRA itself. A smaller but still enormous slice, on the order of $100 million, sat in USDC and SOL that researchers said had already been pulled out of the token’s liquidity pools.
That split matters more than people admit. A bag of a freshly launched meme coin is not the same thing as dollars in a stablecoin. One can go to zero because the market walks away. The other only moves if someone sends it. I’ve found that coverage of these episodes often blends the two, then acts surprised when the headline number collapses even if the dollar-like portion was the real story all along.
The original clustering also split the map into two buckets. One set of wallets was tied more directly to the token. Another set was labeled as the venture entity, for addresses not framed as part of the project itself. Together they produced the near-$300 million print. Separately, they told a more useful story: launch inventory on one side, extracted stablecoins and native gas-token balances on the other.
Why a Cluster Is Not a Bank Account
Attribution in public blockchains is a craft, not a stamp from a court clerk. Analysts group addresses because they share funding paths, timing, signing patterns, or later transfers into known venues. That work can be excellent. It can also be revised. A cluster of 1,000-plus addresses is a research product. It is not, by itself, a finding that every satoshi or lamport inside it was personally controlled by one executive on a given Tuesday.
Recent federal court language describes Kelsier Labs LLC, operating as Kelsier Ventures, as a family-run firm incorporated in Texas in 2021, with Hayden Davis as chief executive. That corporate sketch is useful context. It does not turn every labeled wallet into a personal checking account. Perhaps the most interesting aspect of the October 2026 update is how little it claims. The analytics post did not walk through each transfer. It simply said the tracked set now holds about $2 million.
A falling entity balance is a clue, not a confession. Price, transfers, sales, pool moves, and changes in what gets tracked can all move the same number.
So when someone writes that the stash “vanished,” I get uneasy. Vanished from the labeled cluster is not the same as vanished from the chain, and vanished from the chain is not the same as landed in a personal pocket. Those are three different sentences. Only one of them is supported by the public update.
What the October Update Actually Said
On October 5, 2026, the same analytics shop that first published the full address set reminded the market of its February work. The reminder was short. More than 1,000 addresses. Almost $300 million then, mostly the launch token. About $2 million now. A chart. No transaction ledger attached.
That restraint is easy to miss if you only screenshot the ratio. A drop from roughly $300 million to roughly $2 million is a 99 percent haircut in marked value. It is also compatible with several ordinary explanations sitting on top of one another. The token could have repriced toward dust. Coins could have moved to addresses outside the original cluster. Stablecoins could have been swapped, sent onward, or caught in freezes that later lifted. Some addresses might have been re-labeled. Any serious reading has to hold all of those possibilities at once.
In my experience, the accounts that age well are the ones that refuse the clean villain montage. The accounts that age badly are the ones that treat a dashboard as a wire transfer receipt.
How the Launch Set the Number in Motion
LIBRA went live on Solana on February 14, 2025. Within a short window, Argentine President Javier Milei posted the contract and promoted an accompanying project framed around the phrase Viva La Libertad. Demand ripped the price higher. Then the price broke. Milei later deleted the post and said he had not known the project’s details. Davis later said he controlled funds connected to the launch while denying that the token was built as a scam.
Those two statements can sit next to each other without resolving the fight. Control of funds is a factual claim about keys and transactions. Intent is a legal claim about what someone meant to do. Markets do not wait for that distinction. They reprice in minutes. Courts take months, sometimes years.
The analytics map landed five days after launch, on February 19. By then the violent part of the chart was already over. What remained was the forensic part: who funded the pools, who removed liquidity, which wallets still held inventory, and which stablecoin balances looked like proceeds rather than unsold hype.
- Launch day concentrated attention, liquidity, and political reach into a few hours.
- The early mark-to-market was dominated by the token itself, not by settled dollars.
- A large stablecoin and SOL slice was described as already removed from pools.
- Public labeling arrived after the collapse, not before it.
If you trade these things, that sequence should feel familiar. Attention arrives. Liquidity looks deep for an afternoon. Early wallets interact with the pool. The chart gives back the entire move. Then the spreadsheets show up. The $300 million figure belongs to that second act, the spreadsheet act, more than to the candle everyone screenshotted.
Token Inventory Versus Extracted Liquidity
Think of the original pile as two drawers. Drawer one held LIBRA. Drawer two held USDC and SOL said to have come out of the pools, plus whatever else the cluster contained. Drawer one can shrink because buyers disappear. Drawer two shrinks only if assets move, get frozen, or get reclassified out of the entity page.
That is why comparing today’s $2 million print with the old $300 million print, as if both were cash, flatters the drama and cheats the math. A large share of the old number was a meme coin marked at a post-launch price that did not survive the week. Once that price died, the entity balance had to fall even if nobody touched a key. The interesting residual is whatever dollar-like value left the pools and whether it stayed inside labeled addresses, moved outward, or ended up restrained by a court or an issuer.
| Snapshot | What was counted | What it does not prove |
| February 2025 cluster | Nearly $300 million across 1,000-plus addresses, mostly LIBRA, with about $100 million in USDC and SOL described as pool withdrawals | That the full mark was realized profit or personal cash |
| May 2025 issuer freeze | About $57 million in USDC at two addresses, under a U.S. court order | That the freeze covered the whole cluster |
| August 2025 Argentine order | 323,275 USDT across two addresses linked to the inquiry | That those coins were the same balances as the U.S. freeze |
| October 2026 cluster | About $2 million still attributed to the tracked set | That $298 million was wired out as profit, recovery, or seized funds |
I keep that table in mind whenever a reply guy treats the ratio as a confession. The rows are related. They are not the same event.
Price Collapse Did a Lot of the Work
LIBRA reached a multibillion-dollar valuation on paper, then gave it back shortly after the presidential post. Paper valuation on a thin pool is a magic trick the industry still falls for. A few million dollars of liquidity can print a market cap that looks institutional. Remove the bid, or remove the liquidity, and the cap was never capital in the first place.
So part of the journey from $300 million to $2 million is just the token doing what failed launches do. Inventory that was marked richly became inventory marked poorly. If the cluster still held a mountain of LIBRA after the crash, the dashboard would have shown a mountain turning into a hill, then a curb. No heist required for that leg of the decline. Just a market that left.
Does that mean nothing was extracted? No. Researchers were explicit that a large stablecoin and SOL position looked like liquidity already taken out. Extraction and repricing can both be true. The mistake is using one to erase the other.
Transfers the Update Refused to Narrate
The October note did not say $298 million left as cash. It did not call the change realized profit. It did not call it investor recovery. It did not call it seized funds. That silence is the responsible part of the post, and also the part most likely to be ignored.
Between February 2025 and October 2026, a year and a half of chain time passed. Addresses can fan out. Mixers are less relevant on some networks than bridging, centralized exchange deposits, and fresh wallets funded in small clips. An entity page that stays fixed will miss outbound hops. An entity page that keeps absorbing new addresses can also change the total for reasons that have nothing to do with a single afternoon in February.
Other analytics shops spent those months on the same maze. Some focused on trader profit and loss after the collapse. Others stayed on early wallets and liquidity providers. Attribution improved in places and frayed in others, which is what you would expect once devices, exchange records, and court orders enter the picture. Public labels are a layer. They are not the whole investigation.
The U.S. Civil Case Closed Before the Wallet Update
Less than a week before the $2 million figure recirculated, a federal judge in New York ended the main U.S. investor class action. On September 29, 2026, Judge Jennifer L. Rochon dismissed the amended complaint in Hurlock v. Kelsier Ventures with prejudice. Plaintiffs were not given another chance to rewrite it. The clerk was told to close the case.
The complaint had named Kelsier, members of the Davis family, former Meteora chief executive Benjamin Chow, and other defendants. The theories ran through fraud, conspiracy, the Racketeer Influenced and Corrupt Organizations Act, New York consumer claims, and unjust enrichment. For the Kelsier and Davis defendants, the RICO counts failed to state a claim. Other claims against them failed on personal jurisdiction. Claims against Chow failed to state a claim. Dynamic Labs won dismissal of the claims aimed at Meteora.
Read that carefully. A dismissal with prejudice is a hard stop for that complaint in that court. It is not a verdict that every disputed event around the launch did not happen. The judge decided whether the allegations, as pleaded, met the legal and jurisdictional tests required to keep going. Plaintiffs had started in New York state court in March 2025. The case moved to federal court in May. By late September 2026, that road was shut.
Ending a civil case is not the same thing as writing the chain’s history. It means this set of claims, in this court, will not be retried on a fresh complaint.
Reading of the September 29, 2026 order
I’ve watched readers treat that order as a clean bill of health, and I’ve watched other readers treat it as proof the system looked away. Both reactions outrun the text. Rochon rejected the RICO and fraud theories as pleaded and found jurisdiction lacking for other counts against the venture defendants. That is a procedural and pleading outcome with real consequences. It is not a documentary about wallets.
Frozen Dollars, Lifted Restraints
Wallet tracing was already inside the litigation long before anyone posted a $2 million update. In May 2025, Circle froze about $57 million in USDC held at addresses tied to the dispute, after a U.S. court order. The freeze covered two addresses while investors pressed claims against Kelsier and related defendants. Later in 2025, the New York federal court lifted those asset restraints. The September 2026 dismissal then closed the case underneath them.
Do not line the $57 million freeze up against the $2 million cluster and pretend you have solved for the missing money. The freeze hit two addresses. The cluster covered more than a thousand. The February estimate mixed token inventory, stablecoins, SOL, and other balances. A restraint that is later lifted also does not tell you where coins went after the handcuffs came off. It tells you the court stopped holding them in place for that proceeding.
There is a human temptation here, and I get it. People who bought the top want a single vault with a single key and a single villain. The filings describe something patchier: issuer freezes on specific stablecoin addresses, later relief from those restraints, and a case that ended without a damages award.
Argentina Did Not Close Its File
The U.S. dismissal left a separate criminal inquiry in Argentina untouched. In March, the federal prosecutor’s office said Eduardo Taiano was still running the investigation and still working evidence connected to the token. Prosecutors have been looking at how the coin was created and launched, including possible fraud, influence trafficking, bribery, and abuse of authority. Those are allegations under inquiry, not findings recited here as proven facts.
Digital assets were already frozen on that side of the ocean. In August 2025, prosecutors said an Argentine federal judge ordered a freeze of 323,275 USDT across two addresses identified in the case. Those addresses, prosecutors said, had received funds from a multisignature wallet under examination. Activity in that wallet was linked to the stretch from December 2024 through February 2025. A separate prosecution request had sought to freeze nearly 44.6 million USDC connected to the inquiry, before the U.S. court froze related stablecoin addresses in May 2025.
The technical work was not only on-chain. Prosecutors said seized devices were analyzed. A final forensic report reached them in January 2026 and was added to the case system in February. None of that is resolved by a New York pleading dismissal. Different country, different standard, different file.
- U.S. civil claims ended with prejudice on September 29, 2026.
- Asset restraints in that U.S. case had already been lifted in 2025.
- Argentine prosecutors kept a criminal inquiry open, with prior freezes on identified stablecoin addresses.
- Device forensics were delivered into that file in early 2026.
If you are trying to track the money rather than the meme, this is the fork in the road. One proceeding stopped. Another did not. The public cluster balance is a third thing entirely, a live label that can shrink without either court announcing a recovery.
What $2 Million Still On-Chain Does and Does Not Mean
The analytics firm still groups wallets it attributes to the venture entity, so anyone can watch holdings and transfers that remain inside the label. It has not said the $2 million is everything Davis or the firm controls outside those addresses. The February research was limited in the same way. First to publish a large set is not the same as last to find every wallet.
That caveat should be printed on the chart. Unidentified wallets, exchange accounts, assets moved to chains the label does not cover, and coins already spent are outside the picture by definition. A small residual can mean the cluster was drained. It can also mean the valuable part left the labeled set months ago, while dust and illiquid tokens stayed behind to keep the page from hitting zero.
Would I treat $2 million as “what they have left”? No. I would treat it as what this particular map still sees. Those are different claims, and the second one is the only one the update supports.
How Holders Should Read a Balance Drop
If you held the token, the emotional math is simple and unfair. You remember a number with many zeros. You see a new number with almost none. The gap feels personal. The chain does not do personal. It does inputs and outputs.
A practical reading looks like this. First, separate dead token inventory from stablecoin balances. Second, ask whether a freeze ever touched the specific asset you care about, and whether that freeze is still in force. Third, notice which court is talking. A dismissed civil complaint in New York does not pay you, and it does not automatically release coins frozen under another sovereign’s order. Fourth, assume labels lag reality in both directions.
A cleaner checklist: Mark versus cash Frozen versus moved This court versus that court Labeled wallets versus all wallets Allegation versus finding
None of that returns a loss. It does stop you from building a theory that cannot survive the next filing. I’ve found that the holders who stay sane in these stories are the ones who keep a timeline, not a slogan.
Political Reach Changed the Speed, Not the Mechanics
Most failed launches never earn a presidential post. This one did, briefly, and that is why the liquidity arrived so fast and why the political blowback arrived just as fast. The mechanics underneath were still the mechanics of a pool on a fast chain. Someone supplied inventory. Someone could remove it. Price was a function of what remained against market orders.
Milei’s deletion and his statement that he had not known the details became part of the public record cited in later proceedings. They do not, on their own, tell you who held the withdrawal keys in the first hour. Political attention is a distribution channel. It is not a custody solution. Anyone who bought because a head of state posted a contract was taking a risk that no dashboard can launder after the fact.
There is a lesson here that sounds obvious until you watch the next cycle ignore it. A famous account can move price. It cannot make a pool deep, a team accountable, or a withdrawal reversible. The chain will still settle whoever signed.
Family Firm, Public Label, Private Keys
Court papers describe a Texas company from 2021, family-run, with Davis as chief executive. Public analytics describe a large address set. Between those two descriptions sits the thing neither a filing nor a dashboard fully captures: who could sign, who approved a transfer, and who merely received a downstream payment. Family firms make that question sharper, not easier, because relatives can appear in complaints without sharing a seed phrase.
The dismissed complaint named family members among the defendants. Jurisdiction and pleading rules knocked claims out. That outcome limits what U.S. investors can do in that case. It does not publish a key ceremony. If you want key control, you are back on the chain and, where it exists, on device forensics in the Argentine file.
I would be careful with language that treats every relative named in a complaint as a co-signer. Complaints cast wide nets. Orders that dismiss them are not character references, but they are also not convictions. Stay with what was alleged, what was dismissed, and what is still being investigated.
Why Meteora and Chow Fell Out of the U.S. Case
The launch did not happen in a vacuum. Infrastructure around Solana liquidity, including Meteora, became part of the civil narrative. Claims against Chow were dismissed for failure to state a claim. Dynamic Labs obtained dismissal of claims brought against Meteora. Again, that is a pleading result. It means the court would not let those counts proceed on the amended complaint, not that every integration choice around the pool was reviewed on a full evidentiary record.
For readers trying to reconstruct the stash, the infrastructure fight is a side door. Pool design can make extraction easier or harder. It does not, by itself, identify the later hop of a stablecoin. Once USDC or SOL left the pool, the story became a transfer story. Interface blame and transfer blame are related, and lawyers will always try to stitch them. The chain still shows signatures, not slogans.
Stablecoin Issuers as the Surprise Choke Point
The May 2025 freeze is the clearest reminder that dollar tokens are not bearer cash in every setting. An issuer that can blacklist an address can interrupt a balance that traders treat as settled. About $57 million in USDC stopped where it sat because a court told the issuer to act. That power cuts both ways. It can preserve a pot for litigation. It can also create a false sense that the whole episode was contained, when the order only named specific addresses.
USDT showed up on the Argentine side, in a much smaller freeze of 323,275 units. Different issuer, different judge, different scale. A prosecution request aimed at nearly 44.6 million USDC never became the same thing as a completed global seizure. Requests, orders, and actual blacklists are three layers. Mixing them is how rumor outruns the docket.
When restraints lift, the choke point opens again unless another order replaces them. Coins that were stuck can move. Coins that were never stuck were never in that story. The October cluster balance cannot distinguish those paths for you unless you open the transactions yourself.
A Timeline Worth Keeping
Dates discipline a story that otherwise turns into folklore. December 2024 through February 2025 is the window prosecutors tied to multisig activity under review. February 14, 2025 is launch day. February 19 is the first full public address map. March 2025 is the original class filing. May 2025 brings removal to federal court and the USDC freeze. August 2025 brings the Argentine USDT freeze. January and February 2026 bring the device report into the Argentine system. September 29, 2026 closes the U.S. civil case. October 5, 2026 is the cluster update at about $2 million.
Lay those points out and the $300 million figure starts to look like a snapshot taken near the beginning, not a balance that sat untouched until someone noticed. Eighteen months of motions, freezes, lifted restraints, and forensic work sit between the two prints. Of course the dashboard moved. The only live question is which mechanism moved which slice.
What Would Actually Explain the Gap
Four mechanisms can produce a drop like this, and they can stack. Repricing of unsold LIBRA is the least cinematic and probably the largest in percentage terms. Outbound transfers to unlabeled addresses would shrink the entity even if value still existed somewhere. Sales into markets or deposits to exchanges would convert inventory into something the label might lose. Legal restraints and later releases would park value, then let it leave, without the entity page narrating the court order.
A fifth mechanism is methodological. If analysts drop addresses, the total falls. If they add spent addresses, the total can fall too, because empty wallets dilute nothing but also do not replenish the sum. The October post did not say the methodology changed. It also did not publish a reconciliation. Until someone does, the gap stays underdetermined.
Underdetermined is an unsatisfying word. It is still the right one. I would rather leave a number unexplained than invent a wire that nobody has shown.
Investor Recovery Is a Separate Question
Nothing in the public update identifies the balance change as money returned to buyers. The U.S. class case ended without a path to a judgment in that complaint. Argentine freezes, where they remain in force, are preservative steps inside an investigation, not a distribution plan. If a recovery ever happens, it will arrive through a specific order, a settlement, or a liquidation process with a claims list. A smaller entity balance is not that process.
This is where wishful reading does the most damage. A holder sees $300 million become $2 million and imagines the difference sitting in a fund waiting for claimants. The documents we have do not say that. They say a labeled set is smaller, a U.S. case is closed, and another country’s prosecutors are still working. Hope is not a custody model.
Denials, Control, and the Word Scam
Davis acknowledged control of funds connected to the project and denied that the launch was meant as a scam. Control is the part analytics and, potentially, device seizures can test. Intent is the part criminal law argues over, with standards that differ by country. Using “scam” as a casual synonym for “sharp loss” collapses that distinction. Sharp losses happen in fair markets and in ugly ones. The label should follow evidence, not the size of the candle.
The U.S. plaintiffs tried to plead fraud and racketeering and did not get past dismissal. That does not crown the denial. It means those counts were not adequately pleaded, or the court would not exercise jurisdiction, on the record it had. Argentine prosecutors are still examining fraud and related offenses as possibilities. Both facts can be stated without picking a winner in a paragraph.
If you write about this for a living, the discipline is boring and worth it. Say who claims control. Say who denies bad intent. Say which court stopped which claim. Say which coins were frozen, where, and whether the freeze still stands. Leave the rest marked as open.
What On-Chain Readers Can Still Check
You do not need a subpoena to watch the residual cluster. Public entity pages show token balances and transfers for addresses the firm is willing to name. What you cannot see is off-exchange custody, unidentified wallets, or the contents of a seized laptop. A useful habit is to note the block time of large stablecoin moves in February 2025, then see whether those same coins appear in later freeze announcements. Sometimes they do. Sometimes the announcement is about a cousin address two hops away.
Nansen and other research desks tracked how traders, not just insiders, made and lost money in the collapse. That work is a reminder that the pain was not confined to one cluster. Late buyers funded early exits in the usual way. Mapping the venture wallets explains a slice of supply. It does not explain every red candle.
Would a full transaction-by-transaction reconciliation change the public mood? Probably. It would also be long, technical, and easier to ignore than a ratio. Until it exists, the ratio is a headline, not a ledger.
Risk Lessons That Outlive This Token
Political endorsement is not disclosure. A contract address in a social post is not an audit. Pool liquidity can leave faster than a timeline can refresh. Stablecoins can be frozen by issuers under court order, which is protection in one frame and concentration risk in another. Civil claims can die on jurisdiction even when the chart looks outrageous. Criminal files can outlive those claims by years.
None of this is unique to one ticker. It is the shape of fast-chain launches that borrow attention from outside crypto and then settle inside crypto’s own rules. The $300 million stash, as marketed in memory, was mostly a marked bag plus a large extracted slice. The $2 million stash, as labeled now, is a remainder. The distance between them is the whole story, and it has not been fully told.
If I had to bet on what ages well, it would be this. The number was real as a snapshot and misleading as a fortune. Courts have already split the aftermath into a closed American civil case and an open Argentine criminal inquiry. Freezes happened, and at least the American ones were later lifted. Anyone who tells you the missing millions are sitting in one obvious place is selling a simpler movie than the chain, and the docket, will support.
A Residual Balance Is Still a Live File
About $2 million is not nothing. It is small next to the original print, and it is large next to a typical retail wallet. As long as those addresses move, the label remains a monitoring tool. A quiet cluster can become a loud one with a single outbound transfer. That is the modest usefulness of the October update. Not a solved mystery. A smaller map, still public, still incomplete.
So what happened to the stash? Part of it was never cash. Part of it was described as liquidity already removed. Part of it passed through freezes that did not cover the whole map. Part of it sits outside any label we have been shown. And part of the argument has moved from price charts to court files that do not share a finish line. That is less satisfying than a vanishing act. It is closer to what the records actually allow.
