ai16z Trades: The Dead AI Fund Whose Token Still Moves

Two years ago this month, a Solana token called ai16z turned a recycled venture-capital joke into one of the loudest trades in crypto. An AI agent, the pitch went, would run a hedge fund in public, take tips from its community, and compound them into real money. By the first week of January 2025 the token behind it was worth about $2.39 billion. By August 2026 its founder stood on X and called it dead.

And then, in the first half of October 2026, the token did something a dead thing is not supposed to do. It traded, and it traded hard. ai16z, now carried under the ticker ELIZAOS, roughly doubled in a little over a week, jumping 34.6% in a single session on 7 October to about $0.000566 according to Pintu’s market desk, before giving the whole move back and breaking lower. As of 28 October it changes hands near $0.0002447, down 42.5% on the week and 45.4% on the month, CoinGecko shows. There is no fund behind it now, no foundation, no founder holding a bag, no buyback coming. It trades anyway.

The trade that refuses to end

“ai16z trades” began as a promise and became a punchline, and it is now something stranger than either: a reflex. The project that was sold as the first autonomous AI hedge fund never really traded on its own, the fund it lived inside wound down long ago, and the person who built it says he owns none of the tokens left over. Yet the ticker still prints green and red candles every day, driven by people betting on a logo whose owner has walked away from it.

This is the story of what ai16z actually traded, what its token trades on now, and why the gap between those two things is the most honest lesson the 2024 to 2025 AI-agent mania left behind. It is also, in miniature, a story about how a crypto market prices a brand after the people and the plumbing behind it are gone.

What ai16z trades was supposed to mean

In October 2024 the developer Shaw Walters launched ai16z on Solana through daos.fun, a launchpad that lets a trader raise a pool of capital and run it in public. The name was a wink at Andreessen Horowitz (a16z), the venture firm; the conceit was that an AI agent, not a human, would allocate the money. The target raise was modest, a small five-figure sum.

The structure matters, because it shaped everything that followed. On daos.fun, a creator has roughly a week to hit a funding goal. If the raise succeeds, the trader controls the treasury and the fund’s token trades on an automated market maker, its price floating with perceived performance. At expiry, holders can burn their tokens for a proportional share of the fund’s holdings or SOL, with the proceeds split between the creator and investors. In other words, the token was always a claim on a pot of money run by one person, dressed as a claim on a machine.

What turned a modest experiment into a mania was a single tweet. On 27 October 2024, Marc Andreessen posted “GAUNTLET THROWN” alongside the bot’s profile, then added “Hey, I have that T-shirt.” The market cap rocketed to roughly $96.6 million within hours, and the attention briefly crashed the daos.fun site. ai16z ran two on-chain personas, a measured “Marc AIndreessen” (handle pmairca) and an aggressive “Degen Spartan AI,” and marketed a “Marketplace of Trust” in which community members pitched trades and gained influence as their calls proved profitable.

The fund that never really traded itself

The problem sat in plain sight from the start. Within days of the Andreessen bump, Protos reported that the breakthrough was human-operated. “Of course, no fund on Daos.fun is actually operated by an AI,” the outlet wrote, describing the creator as “a non-AI human named Shaw” and the practice as “AI-washing,” the habit of spraying a project with a thin coat of machine-intelligence credibility. “In reality, humans trade meme coins with users’ funds instead,” it concluded.

That is the crux of the whole saga. The ai16z trades people cheered were, as far as anyone can verify on-chain, discretionary human trades wrapped in an AI narrative: positions in SOL, in the project’s own token, and in a rotating set of correlated memecoins. There was never a published, independently auditable track record showing an algorithm opening and sizing positions on its own. The agent was real software, but the money decisions were not the software’s to make.

The “Marketplace of Trust” framing is where the human and the machine blurred on purpose. In the pitch, community members submitted trade ideas, the agent weighed them against each contributor’s track record, and good callers earned more influence over time. It sounded like a self-improving collective intelligence. In practice it was a feedback form with a mascot: a human still decided what to act on, and the trust score was a reputation layer, not a trading engine. That distinction matters because the whole investment case rested on the idea that the allocation was automated, and therefore tireless, unemotional, and scalable. None of those properties hold if a person is in the loop on every position.

Walters himself later said the quiet part out loud. Speaking at Token2049 in Singapore in October 2025, he told Decrypt: “You probably do not want to give an AI agent a bunch of money and expect it to make you more.” Coming from the person who built the most famous AI-fund token in the market, it was about as close to a retraction of the original pitch as crypto gets.

From $96 million in an afternoon to $2.4 billion in a quarter

Momentum did the rest. The token’s price peaked at an all-time high of $2.47 on 1 January 2025, and its market cap topped out near $2.39 billion a day later, CoinDesk reported. For a few weeks, a repackaged venture-capital in-joke was worth more than many publicly listed asset managers.

Then the branding caught up with it. In late January 2025 the project rebranded from ai16z to ElizaOS, after Andreessen Horowitz asked for distance from its name. The token kept the AI16Z ticker at first, then migrated to ELIZAOS on a snapshot around 11 November 2025, a process that lifted the supply target toward 11 billion tokens. The framework and the fund, once the same story, began to pull apart: one was open-source code, the other a speculative asset with a shrinking reason to exist.

The decline was not a single crash so much as a long bleed punctuated by the occasional violent bounce, the pattern of a low-float token with no cash flows and a story that kept changing underneath it.

Two ghost tokens, one graveyard

Here is a detail that still trips up newcomers: there are two ai16z tokens, and both still trade. The migrated token, ELIZAOS, carries almost all of the remaining activity. The unmigrated legacy AI16Z contract lingers on Solana with a market cap under $400,000 and daily volume that some sessions measure in the low four figures, per CoinGecko. They move independently, which is itself revealing: in the October run-up, speculators piled into the migrated ticker while the legacy contract barely twitched.

Add the two together and the surviving market value of the entire ai16z complex is around $2.2 million, down roughly 99.9% from the $2.39 billion peak. The all-time high in the ELIZAOS column, $0.01078, is the migration-day launch reference rather than a mania-era print; the mania-era high sits on the legacy AI16Z line at $2.47, which the exchange tracker now shows as a near-total loss from the top.

Anatomy of the October 2026 zombie rally

So what actually moved in October? Not fundamentals. The spike that took ELIZAOS up 34.6% on 7 October, and more than doubled it over the following week and a half, had no product launch, no treasury, and no founder endorsement attached to it. Walters has repeatedly said he holds none of the token and will not support it. The most grounded reading, and the one the price action since has confirmed, is a thin-liquidity move: a low-float token with a famous name, caught in a broader rotation into small-cap AI tokens, amplified by a news cycle around Walters’ new venture that had nothing mechanically to do with the old token.

Thin liquidity cuts both ways. With only a couple of hundred thousand dollars of daily turnover, it takes very little buying to lift the price and very little selling to collapse it. That is why the round trip was so complete: by 28 October the token sat below where the rally began, nursing a 45.4% monthly loss even after the pump. A chart like that is not a verdict on any technology. It is a verdict on float, attention, and the reflexive way crypto trades names it recognizes.

ai16z is not the only token to behave this way after its story ended. Crypto is full of “zombie” tickers, projects whose teams have moved on but whose tokens keep a listing, a chart, and a small crowd of traders who treat the name as a tradeable meme. The pattern tends to rhyme: dormant price, a sudden burst of volume on a narrative or a sympathetic move elsewhere in the sector, a vertical candle, then a slow fade as the handful of buyers run out. What makes ai16z a cleaner case study than most is that its founder has been unusually explicit that there is nothing to buy, which removes any ambiguity about what the October move was and was not.

The irony is hard to miss. The one thing ai16z was marketed to be, a disciplined allocator, is the exact opposite of how its token now behaves. The ticker trades on momentum and liquidity mechanics, the two forces the Marketplace of Trust was supposed to replace.

On-chain and transparent, off the books and unauditable

Crypto’s favorite defense of projects like this is that everything is on-chain, so anyone can check the trades. That is true and beside the point. You can see every transfer the ai16z and DAO wallets ever made. What you cannot reconstruct from the chain is intent: which transfers were the agent’s decision, which were a human’s, and which were market-making, treasury management, or insider positioning. Transparency of transactions is not the same as auditability of a strategy.

This is the paradox that makes ai16z trades such a slippery phrase. A regulated fund publishes audited returns, a methodology, and a custodian’s records. ai16z published a wallet address and a narrative. The gap between those two is precisely where the dispute later lived, and it is why a public ledger did nothing to settle the central question of whether the fund ever did what it claimed.

For anyone sizing up the next on-chain AI fund, the takeaway is blunt: a visible wallet is evidence of activity, not evidence of skill, and certainly not evidence that software rather than a person is pulling the trigger.

What would real auditability have looked like? At a minimum, a published mapping of which wallet made which decision and why, timestamped signatures proving the agent rather than an operator initiated each trade, and a performance report reconciled against the treasury’s actual holdings. On-chain forensics firms can approximate some of this after the fact, but they are reverse-engineering intent from transfers, which is guesswork dressed as rigor. The project never shipped the primitives that would have made its central claim falsifiable, and in crypto, a claim you cannot disprove is usually one you should not pay for.

The daos.fun clock that never rang

daos.fun funds are not meant to run forever. They carry a set term, and at expiry the mechanism is clean: the treasury freezes, holders burn tokens for their share of the remaining assets or SOL, and the pool is split between creator and investors, per the platform’s own description. For a fund created in late October 2024, that clock pointed at roughly the end of 2025.

It never rang on schedule. Instead of a tidy wind-down at the one-year mark, ai16z became ElizaOS, the token migrated, the supply target ballooned, and a clean burn-for-assets settlement was replaced by an open-ended token that simply kept trading. The structure that was supposed to give holders a defined exit became a treadmill with no off switch, which is a large part of why the grievances that followed were about expectations as much as about money.

When the real wind-down finally came in 2026, it did not come through daos.fun’s burn mechanism. It came through a courtroom.

The 60/40 split, the lawsuit, and a judge who wanted one more try

On 20 April 2026, Burwick Law filed a class action, Pikabea v. Walters, in the Southern District of New York (case 1:26-cv-03238). The docket names Eliza Labs, Shaw Walters, ElizaOS director Sebastian Quinn-Watson, and the AI16Z DAO. The complaint pleads New York and California consumer-protection and false-advertising claims plus unjust enrichment, alleging the project misused the a16z name, overstated the agent’s autonomy from the Protos report onward, and structured the token migration to favor insiders.

That last allegation is the one that connects ai16z to a pattern crypto keeps repeating. The suit describes a migration split of roughly 60% to existing holders and 40% to insiders, and identifies thousands of wallets, upward of 3,945, carrying net losses over the class period. Whether or not those exact figures survive discovery, the shape is familiar: a community token whose economics tilt toward the people closest to it. Readers who followed our look at gaming DAOs and governance capture will recognize the dynamic, where the language of community ownership and the distribution of actual control quietly diverge.

The case moved slowly, and for an almost comic reason. At a hearing on 19 May 2026, Judge Jed Rakoff learned that plaintiffs had tried to serve Walters exactly once, at a San Francisco address. “I am not going to move the case forward on this basis: 1 attempt?” he said from the bench, per court reporting from Inner City Press. Serving a defendant in Australia, and a DAO that, as Walters’ own counsel noted, cannot be sued in the usual way, is the kind of friction that makes DAO litigation drag.

Underneath the procedural comedy sat real losses. The holders who bought the AI-fund story near the top, migrated their tokens in good faith, and held through the rebrand are the people the class action was built to represent. By the time the settlement arrived, their tokens were worth a fraction of a cent, and the recovery was a share of a drained treasury rather than anything resembling the original outlay. That is the uncomfortable footnote to every headline number: the $2.39 billion was never evenly distributed on the way up, and the collapse was not evenly shared on the way down.

“The token is dead. Completely.”

The resolution arrived not as a verdict but as a surrender. On 4 August 2026, Walters announced on X that the foundation had transferred its remaining treasury and available funds to settle with the Burwick group, and that the token was finished. “The token is dead. Completely,” he said, adding there would be “no further foundation support, buybacks or supply measures,” CoinDesk reported. He framed the settlement as a capitulation rather than an admission, saying the claim “was ridiculous, but we didn’t have the capital to legally fight it.” He also said he once held tokens worth about $25 million and watched them fall toward zero.

Crucially, Walters did not stop building. He said he would keep developing the ElizaOS software without launching another token, and that he was “starting over, since I own the IP, and I am never letting a token come close to Eliza again.” The open-source framework remains active, with roughly 19,600 stars on GitHub, a reminder that the code and the coin were always separable.

His next move is the most telling part. In October 2026, Walters’ new project, Slop Cash, was named the first external launch on Umia, a platform that raised $6.11 million at an $18 million valuation to help projects launch and govern tokens through legal wrappers and on-chain auctions. Slop Cash pays open-source contributors in stablecoins for merged work, keeping the framework itself tokenless. Umia chief executive Francesco Mosterts described the design as a way to ensure “value generated by a project goes to the token rather than remaining with founders, equity holders or a separate foundation.” The tokens, in other words, did not vanish from Walters’ world. They moved one layer up, into the funding and governance stack, where he argues they belong.

What ai16z proved about trusting an AI with money

Strip away the lawsuit and the price chart and ai16z leaves one durable lesson: in 2024 and 2025, the market was willing to pay billions for the idea of an autonomous trading agent long before the technology could deliver one safely. The autonomy that would have justified the valuation is exactly the autonomy that makes an agent dangerous to fund.

ai16z did not rise in a vacuum. It was the flagship of a late-2024 wave of AI-agent tokens, alongside experiments like Truth Terminal and a crowd of autonomous posting-and-trading bots, that convinced a lot of people the next cycle would be run by software. That flagship status is exactly why its unwinding carries weight. If the most famous, most endorsed, and best-funded AI fund in the market turned out to be a human trading memecoins behind a chatbot, the burden of proof on every imitator rose accordingly.

Ethereum co-founder Vitalik Buterin spent much of 2026 arguing that the safe way to run agents is defensive and local. In an April 2026 post on his own security setup, he urged builders to keep “all LLM inference local first,” to host files locally, and to “sandbox everything,” warning readers not to assume any off-the-shelf toolchain is secure. An agent you can trust with a wallet is, by his logic, a heavily constrained one, which is the opposite of the free-roaming genius ai16z was sold as.

The contrast with where on-chain AI trading actually landed is instructive. The 2026 version that works is narrow and plumbed into intent-based systems and solver networks, where an agent states a goal and specialized actors compete to execute it, often inside vaults with hard spending limits. It is less glamorous than a bot that runs a hedge fund, and far more honest about what software can and cannot be trusted to do with money. ai16z was the expensive public demonstration that the glamorous version was a story, not a product.

Where a dead token still trades, and what the trade costs

If you want to trade ai16z today, you are mostly trading ELIZAOS on a handful of venues that list long-tail tokens, the smaller centralized exchanges and the Solana decentralized exchanges where thin order books are the norm. Access is not the hard part; cost is. On a token turning over a couple of hundred thousand dollars a day, spreads are wide, slippage on any meaningful size is real, and the quoted price can be a poor guide to the price you actually get. The same arithmetic we walked through in what a perp DEX trade really costs applies with force here: the headline number is rarely the number that clears.

Listing status is its own risk. Dead or dying tokens are prime candidates for delisting as exchanges prune inactive markets and change their promotions, a reminder that where you can trade something is not fixed; our coverage of the shifting terms at CoinEx is one small example of how quickly venue rules move. Many holders of long-tail alts first found them through sign-up and referral promotions on exchanges like MEXC, which is a reasonable on-ramp but does nothing to change the underlying liquidity of a token nobody is steering.

Scale is the last piece of context. Even after October’s pump, ELIZAOS is a rounding error next to the sector it helped popularize.

The broader AI Agents category is worth about $3.44 billion, per CoinGecko. ELIZAOS is less than a thousandth of it. The name that once defined the trade is now a footnote to it.

For an individual still holding, the math is unsentimental. A position bought near the peak is now a near-total loss on paper, and in many cases the dollar value is too small to be worth the gas or the trading fees to exit. Some holders keep the tokens as a souvenir of the cycle; others will realize the loss for tax purposes. Either way, the choice has nothing to do with the project, because there is no project left to have a view on, only a ticker and a thin market.

The SEC never showed up

For all the money involved, the reckoning for ai16z came from private plaintiffs and from the market, not from Washington. The Securities and Exchange Commission never brought a case. That absence reflects both the agency’s lighter-touch posture toward crypto in 2026 and a genuine gap in the rulebook: an AI agent has no legal personhood, so when an “autonomous” fund misleads, liability falls on the humans and companies behind it, which is exactly where the Burwick suit aimed.

SEC Chair Paul Atkins has described his philosophy as setting boundaries rather than steering outcomes. The job, he said in May 2026 remarks, is “to set the rules of play and referee the game, not to pick the winning team.” For a token like ai16z, that means the main guardrails against the next AI-fund fantasy are disclosure rules, anti-fraud law, and private litigation, not a bespoke AI-agent regime. Buyers of the ghost token today enjoy no special protection; they are trading a disavowed asset at their own risk.

What to watch from here

The honest base case for the ai16z tokens is slow irrelevance: thinning volume, the occasional attention-driven spike like October’s, and a standing risk of delisting. Nothing about the project’s current structure generates cash flow or demand, so any future rally will be, like the last one, a liquidity event rather than a fundamental one.

The more interesting thread is Walters himself. If Slop Cash and Umia work, they become the real post-mortem on ai16z, a deliberate attempt to put tokens where incentives belong and to pay builders for shipped code rather than for hype. If they stumble, the lesson of ai16z, that a compelling AI story can outrun both the technology and the governance for a long time, will simply be waiting to be relearned by the next project. Either way, the trade that began with “GAUNTLET THROWN” is effectively over. The candles that still print are an echo, not a fund.

Frequently Asked Questions

Is ai16z dead, and can you still trade it?

The project’s founder, Shaw Walters, declared the token dead in August 2026 after settling a class action, and the foundation has wound down. The tokens, however, still trade: the migrated ELIZAOS and the legacy AI16Z both change hands on Solana, with a combined market cap around $2.2 million as of 28 October 2026, down about 99.9% from the 2025 peak.

What is the difference between ai16z and ELIZAOS?

They are the same project at different stages. ai16z launched in October 2024 and rebranded to ElizaOS in January 2025; holders then migrated from the AI16Z ticker to ELIZAOS around November 2025. A legacy AI16Z contract still trades for those who never migrated, but ELIZAOS carries nearly all remaining activity.

Did the ai16z AI agent actually trade on its own?

There is no independently auditable record that it did. Protos reported within days of launch that the fund was human-operated, and Walters later said you probably should not expect an AI agent to grow your money. The trades were discretionary positions wrapped in an AI narrative, not a verifiable autonomous strategy.

Why did ELIZAOS pump in October 2026 if the project is dead?

The move had no fundamental trigger. It was a thin-liquidity, low-float rally fed by rotation into small-cap AI tokens and a news cycle around Walters’ new venture, with no mechanical link to the old token. The price round-tripped the entire gain within the month, ending October down about 45%.

What happened to the ai16z lawsuit and Shaw Walters?

Burwick Law’s class action (Pikabea v. Walters, SDNY) was settled in August 2026 when the foundation handed its remaining treasury to the plaintiff group. Walters says he owns none of the token and has moved on to Slop Cash, a tokenless open-source funding project launching through the Umia platform.

By Marcus Okafor, senior markets writer at HOGE Wire, covering AI-crypto, market structure, and the places where the two collide.