Eliza OS founder Shaw Walters has said the ai16z token is over, the foundation has been shut down, and there will be no buyback for holders.
In a long post cited by the original article, Walters told holders they could either sell their tokens or “find a group of people to pump it,” but should not expect him to do anything else. He also described a legal fight, saying law firms had sued him on behalf of token holders and that the Eliza OS Foundation did not have the money to keep fighting. According to his account, the foundation used its remaining funds for payouts.
Walters added that he once had $25 million worth of ai16z tokens in his wallet, never sold any of them, and watched them go to zero. He said he is now living off savings in a run-down small bedroom in San Francisco and spends his days writing code.
From a 420.69 SOL raise to a $2.6 billion market cap
The article places ai16z in the center of the AI Agent token boom that swept through crypto at the end of 2024. In October 2024, ai16z launched a crowdfunding round on DAOs.fun with a target of 420.69 SOL, roughly $75,000, to build an investment fund managed autonomously by AI.
The number 420.69 was itself a crypto in-joke, and the article argues that the tone of the project was visible from the start. Even so, ai16z took off quickly. In less than three months, its market capitalization reached $2.6 billion, according to the piece. It also helped push an entire category higher, with GOAT before it and VIRTUAL after it, as more Agent-themed tokens flooded into the market. The broader AI Agent sector grew from zero to nearly $10 billion, the article said.
That run happened before AI Agent products had fully arrived in practice. The article notes that by late 2024, ChatGPT had only just turned two years old and was still widely criticized for hallucinations. Claude had not yet introduced tools that could directly operate a computer, and for most people the idea of an “AI Agent” remained conceptual. Crypto, however, had already assigned the concept a price.
Early doubts did little to slow the trade
The article says that less than a week after ai16z launched, a crypto media outlet questioned whether the project’s core AI Agent, the supposed autonomous investor “Marc AIndreessen,” was actually being operated by humans rather than making decisions on its own.
It says similar doubts later surfaced around the well-known market analysis Agent aiIxbt. At the time, though, those concerns did not produce much FUD. The market cap kept rising, the community kept cheering, and more AI Agent tokens kept launching.
The piece treats that episode as a compressed picture of the broader crypto AI narrative of the period: whether the product was fully real mattered less than whether the story was ahead of the market. In an environment of excess liquidity, a narrative with first-mover appeal was enough to trigger a wave of speculation. In that setup, tokens existed before products and prices moved ahead of the technology.
By 2026, AI Agents arrived, but crypto did not capture the payoff
The article then shifts to 2026 and argues that AI Agents have actually arrived. It points to CodeX and Claude outside China, and to products such as Workbuddy for domestic use cases. The functions once written into crypto AI white papers — automated market analysis, workflow handling, sentiment monitoring, and similar tasks — have become real products.
But the groups delivering that reality were not the token projects that led the earlier hype cycle. Instead, the article points to Anthropic, OpenAI, and software teams writing code in San Francisco and Hangzhou. Money has flowed to platforms such as OpenRouter, to cloud providers through GPU bills, and to every API call that actually runs in production.
Walters, in the quoted post, put the contrast bluntly: “AI doesn’t have these problems. The people there are optimistic, powerful, and building the future instead of tearing each other down. Those are my people.” The article frames that line as a sharp conclusion from someone who once helped ignite the crypto AI Agent narrative.
Token models faded, but the open-source framework did not disappear
The article argues that today’s successful AI applications and large model companies do not need tokens. It cites Anthropic’s subscription and API revenue model, and says OpenRouter’s relay-platform business has annual revenue above $100 million. None of those companies issued tokens or built DAOs, yet their business models work and developers effectively vote through usage and invoices.
The contrast with crypto AI projects is harsh. When token prices hold up, the article says, a project can appear to have a treasury, a community, and an active narrative. Once prices break down, all of that can unravel at once. Developers leave because token-based pay has lost most of its purchasing power, communities turn into rights-protection groups, and teams trying to build may find they do not even have the budget to hire one engineer.
At the same time, the article notes that the Eliza framework itself is not dead. Its code is still being updated, and companies are still discussing partnerships around it. As the piece puts it, an open-source AI framework can survive without a token, whether through services, acquisition, or other paths that leave it with more room than the token structure did.
It also highlights one more detail: among the active contributors to Eliza over the past week, the article says, was Claude. On that basis, it argues that in the 2026 market, an AI developer or project without a token attachment may be worth more than one tied to a token.
Crypto and AI have not split completely
The article does not treat ai16z’s collapse as the end of every crypto-AI path. Its argument is narrower: what broke down was the route of issuing tokens around an AI concept, not every possible connection between crypto and AI.
It points to projects that are still operating on a different model. Bittensor is building a decentralized AI compute market, while Render is focused on a distributed GPU rendering network. In the article’s view, those projects still have a market because tokens there are used to coordinate actual computing resources, though their case becomes harder to sell without a bull market premium.
It also lists practical areas where AI is starting to matter inside crypto itself, including on-chain data analysis, vulnerability scanning, whale wallet tracking, sentiment monitoring, and contract strategies for decentralized exchanges. The example it gives is a professional trader opening Hyperliquid in 2026 with a Claude- or GPT-powered strategy window open beside it.
That suggests the relationship has changed direction. Rather than packaging AI as a token narrative, the more durable path may be AI moving directly into crypto trading, research, and infrastructure workflows.
What ai16z’s ending says about the sector
The article closes on a broader industry point. Crypto often sees a direction early, it says, but seeing a direction first does not always mean standing in the right place. For developers, dropping the token burden may open more valuable paths in AI than a token ever could. For crypto participants, actually adopting AI may offer more than the earlier rough-and-ready model built around narrative momentum.
ai16z’s path — from a small crowdfund to a breakout market cap and then to a founder declaring the token dead — marks both the end of a project and the retreat of a particular crypto AI Agent trade. Based on the signals in Walters’ post, the token is done, but Eliza OS and the open-source work around it have not fully stopped.