Author|David, Deep Tide TechFlow
ai16z, do you still remember this name?
If you do, it means you experienced the wave of AI coin frenzy in the crypto space at the end of 2024. If you don't, that's also normal; the category of AI Agent tokens has been quiet for a long time.
Crypto x AI, in this era where AI tools are gradually being implemented and progressing rapidly, seems too much like a pseudo-concept and an old-fashioned toy.
Today, Shaw Walters, the founder of the Eliza OS project, posted a very long article, the gist of which is that the ai16z token is completely dead, the foundation is closed, there will be no buyback, and token holders should handle their tokens as they see fit.
He suggested you either sell them or "find a group of people to pump them up," but don't expect him to do anything more.
The tone of this message doesn't sound like a founder gracefully announcing the end of a project; it's more like someone venting frustration at everyone and then slamming the door shut in disappointment. He called the crypto community "a bunch of spoiled crybabies," saying there was a law firm taking the side of token holders and suing him. The Eliza OS Foundation had no money for a legal battle and spent all its remaining funds.
Shaw also said his wallet once held $25 million worth of ai16z tokens, which he never sold, watching them go to zero; he now lives off his savings, staying in a shabby small bedroom in San Francisco, writing code every day.
From his perspective, perhaps he believed he was developing with idealism in a murky market, only to be left with ruin, disappointment, and anger.
The author does not intend to judge these personal emotions. But standing at the crossroads where hot topics change, industry focus shifts, and AI rises, the shutdown of Eliza OS inevitably leads to reflection—
Crypto x AI truly woke up early but arrived late to the party.
A Step Ahead, Riding the Wave
In October 2024, ai16z launched a crowdfunding campaign on DAOs.fun, aiming to raise 420.69 SOL, roughly equivalent to $75,000, to create an AI-autonomously managed investment fund.
420.69—those familiar with crypto culture know this is a meme number. From the choice of fundraising amount, the tone of the project was already set.
But it really took off.
In less than three months, ai16z's market cap surged to $2.6 billion. It also spawned an entire category, with GOAT leading the way, VIRTUAL following, and various Agent-themed coins emerging one after another. The entire AI Agent sector grew from zero to nearly $10 billion.
Crypto was indeed a step ahead conceptually.
What was the AI landscape like at the end of 2024? ChatGPT had just turned two, often criticized for its hallucinatory answers; Claude didn't yet have tools to directly operate computers, and most people's understanding of "AI Agent" was still at the conceptual level.
Yet the crypto market had already priced this concept.
But there was a detail that few paid attention to at the time.
Less than a week after ai16z launched, a crypto media outlet published a report questioning whether the core AI Agent of the ai16z project—the so-called "Marc AIndreessen" that claimed to make autonomous investment decisions—was actually operated by humans, rather than a real Agent thinking on its own.
This same suspicion later arose regarding the well-known crypto market analysis Agent, aiIxbt.
At the time, such doubts could hardly cause any waves or FUD. The market cap kept rising, the community kept hyping, and new AI Agent tokens kept launching...
Looking back, the author thinks this might be the most precise snapshot of the entire crypto AI narrative.True or false, it didn't matter.
In an environment of liquidity spillover, a narrative about a leading version was enough to spark a wave of speculation.
Concept authentic, direction correct, but tokens existing before products, prices existing before technology—this was precisely the essence of crypto being a step ahead.
The Future Has Arrived, But It's Not For Us
In 2026, AI Agents truly arrived.
In the West, there's CodeX and Claude; in the East, there are competitors like Workbuddy, more suited to the domestic context. Looking back at the narratives that lived only in whitepapers during the crypto AI hype wave—such as helping with automatic market analysis, workflow processing, sentiment monitoring, etc.—they have now become a reality.
This might be one of the few realized narratives in the crypto industry, aside from stablecoins.
But the ones delivering them are a different group of people.
It's Anthropic, it's OpenAI, it's the teams coding in San Francisco and Hangzhou. The money flows to aggregation platforms like OpenRouter, to cloud providers' GPU bills, to every genuinely executed API call.
Another sentence from Shaw's long post is deeply thought-provoking:
"AI doesn't have these problems. The people there are optimistic, powerful, building the future instead of tearing each other down. Those are my people."
The person who once ignited the crypto AI Agent narrative concludes that this industry is incompatible with AI?
Objectively speaking, the current successful AI applications or large models indeed have no reason to issue tokens. Anthropic charges through subscriptions and API calls; OpenRouter's intermediary business has an annual turnover of over a hundred million. These companies never issued tokens, never built DAOs, but their business models work, developers vote with their feet, and bills are the best consensus mechanism.
Conversely, the situation for crypto AI projects is harsh.
When token prices were up, everything seemed present—treasury had money, community had people, narrative had heat. Once prices collapsed, everything vanished simultaneously. Developers left because the purchasing power of token salaries shrank by 90%, communities turned into维权 groups (rights protection groups). Those wanting to do serious work might eventually find that the project couldn't even budget for hiring an engineer.
Ironically, the Eliza framework itself in his hands isn't dead. The code is still being updated, and there are businesses discussing cooperation on it. An open-source AI framework can survive without issuing tokens—selling services, getting acquired, the paths are much wider than with tokens.
Even more sadly, among the active contributors to the Eliza open-source AI framework, which originated in crypto, in the past week was actually Claude.
An AI developer or project not tied to tokens, in the 2026 market, might be worth more than when it was tied to tokens.
Embrace the Change
As someone working in crypto, seeing once-hot projects fall one by one is inevitably lamentable.
But sentimentality doesn't mean complete pessimism.
The crypto AI narrative has indeed shattered into pieces. What shattered is the path of "issuing tokens for AI concepts." The possibilities between crypto and AI are not completely severed.
The few surviving projects are doing things completely different from ai16z. For example, Bittensor is building a decentralized AI compute power marketplace, and Render is creating a distributed GPU rendering network. They all use tokens to coordinate a real computing resource; of course, there's still a market for them, just without the premium and persuasiveness they had during a bull market.
Furthermore, within the circle, on-chain data analysis, vulnerability scanning, whale wallet tracking, sentiment monitoring, DEX contract strategies—AI is gradually becoming effective in these scenarios. In 2026, when a professional crypto trader opens Hyperliquid, there's a high chance a Claude or GPT-supported strategy window is open beside it.
This also means the relationship between crypto and AI is not over,but the direction needs to be reversed.
AI empowering all scenarios has become a kind of political correctness, and crypto can also be covered. As for how big the crypto plate still is, we might have to ask when BTC will finally recover.
On a personal level, what one should do is embrace the change.
A developer who sheds the token burden is worth more in today's AI world than tokens could ever give; and a crypto player who solidly embraces AI can clearly obtain more benefits than the previous rough-and-tumble model.
Being in crypto, we often see the direction a step ahead—or maybe we just reversed our own position.
