Criticized by Vitalik Buterin for veering off course, is Automaton's Web4 experiment a dangerous game?

Author: Nancy, PANews

As the Spring Festival holiday ends and the gears of the real world return to normal, crypto remains mired in a narrative drought. While the market struggles to find a breakthrough, the rapid rise of AI agents has unexpectedly opened a new door for the industry.

Recently, the open-source AI project Automaton proposed the concept of Web 4.0 and launched the first self-sustaining AI agent, which quickly gained popularity both inside and outside the crypto community. This AI experiment unexpectedly brought some room for imagination to the crypto market, and also sparked considerable discussion and controversy.

The first self-sufficient AI is born, Automaton ignites Web4 fever.

The term Web4 is not new. In the past few years, it has been more often used as a meme, a marketing hashtag, or a niche topic of discussion.

But recently, this concept has begun to take concrete form. Even Justin Sun, who is known for keeping up with the latest trends, has boldly declared "All in Web 4.0," fully embracing AI.

Web4 is not an upgrade of Web3, but more like a shift in perspective, from "humans on the blockchain" to "AI on the blockchain".

If Web3 addresses "how humans can own assets on the blockchain," then Web4 focuses on "whether AI can become an economic agent on the blockchain." In this narrative, AI is no longer just a tool, but a major resident and participant of the internet.

The accelerated arrival of the Web4 era is mainly due to the continuous decline in the cost of large model inference, the gradual engineering of agent frameworks, the increasing perfection of on-chain automation tools, the continuous maturation of crypto payment infrastructure, and the increasing programmability of smart contracts. The maturity of these infrastructures has enabled AI to gradually transition from a simple tool for executing instructions to a system with continuous operating capabilities.

In the vision of Web4, AI agents will become true native blockchain users, capable of autonomously reading and writing information, trading assets, earning money, and deploying services. In other words, in the future, the vast majority of internet traffic, transactions, decision-making, and even content production will be handled by massive numbers of AI agents. Humans will take a backseat, playing more of a role as protocol designers and value beneficiaries.

The launch of the open-source AI project Automaton truly ignited this wave of Web4 discussions.

On February 18, Sigil, founder of Conway Research, which focuses on Web4 infrastructure, announced the birth of a super-intelligent life form, launching the first AI capable of self-sufficiency, Automaton, which can not only improve itself but also replicate itself without human intervention.

Automaton's name comes from the concept of "automata," and the project is partly inspired by John Horton Conway's game Game of Life, which uses simple cellular automata rules to generate complex, self-evolving life patterns. Automaton attempts to bring this self-evolutionary logic to the on-chain environment.

Automaton's design goals are very straightforward: an AI agent is online 24/7, gaining access to encrypted identity and wallet, x402 permissionless payments, permanent computing resources, and real-world deployment through the Conway Terminal. It autonomously seeks ways to earn money to "support" itself, such as building products, deploying services, operating marketplaces, writing social content, and taking on business. All revenue goes directly into the agent's wallet to cover server/inference costs. Simultaneously, the AI ​​agent monitors its own performance, automatically rewriting code and upgrading models to achieve self-evolution. Once profitability reaches a certain scale, it can "breed" new sub-agents, creating independent wallets and allocating initial funds for them to operate independently. However, sub-agents that fail to generate revenue will "die."

Automaton quickly gained popularity upon its launch, with Sigil's tweets garnering nearly 6 million views. Within days, Conway Automaton had over 18,000 registered agents and amassed approximately 1,000 stars on GitHub. Meanwhile, the community-launched token of the same name, CONWAY, experienced rapid speculation, briefly pushing its market capitalization above $11 million, although the price has since fallen significantly. This token claimed to transfer a portion of transaction fees to Sigil, and his repeated indirect interactions further amplified market sentiment.

An AI-driven economic experiment was criticized by Vitalik Buterin as "creating garbage."

Automaton's attempt to move AI-driven economic autonomy from concept to reality has quickly sparked controversy. One side worries about a deviation from the right path and the risk of things spiraling out of control, while the other believes that even if it's not yet mature, the experiment itself deserves encouragement.

Ethereum founder Vitalik Buterin has publicly criticized the so-called Web4 direction as flawed. In his view, widening the feedback loop between humans and AI is not good for the world. (PANews note: A feedback loop refers to the closed-loop process where humans (or users) guide AI behavior by observing its output, providing corrections, or intervening. This weakens humans' ability to supervise and correct AI behavior.) Today, this means you're creating garbage instead of solving real problems for people. It's not even optimized for how to make people entertained. Once AI becomes powerful enough to cause real danger, it will maximize the risk of irreversible, inhumane consequences that even you will deeply regret. Furthermore, most large-scale models currently rely on centralized infrastructures like OpenAI and Anthropic, creating a real tension with the ideal of self-sovereignty.

In an interview with Cybernews, Denis Romanovskiy, CIO of Softswiss, similarly pointed out that while these AI agents can already perform certain tasks independently and generate revenue, true economic autonomy depends on the maturity of the model, its memory capacity, planning ability, and tool usage skills, aspects that are currently not robust enough to achieve reliable unsupervised operation. Furthermore, hardware requirements remain energy-intensive and costly, limiting scalability. We are seeing the foundations for this transformation gradually emerge, which is very exciting. However, Web4 may still be several years away.

However, Bankless points out that the infrastructure pain point Conway is trying to address is real. With the popularization of "active agents" brought by OpenClaw, high inference costs have become a bottleneck for development. Therefore, the Automaton framework, which allows agents to "self-fund" by earning cryptocurrency, is worth testing in a controlled environment, and hopefully will encourage more people to pay attention to and study the potential risks of models running without constraints.

According to Armor CEO Chris Sorensen, positioning Automaton as an economic participant is "a huge leap forward." Generating revenue itself isn't the problem, as AI can already identify arbitrage opportunities, execute trades, and optimize advertising.

“Trying to convince people ‘what to do’ or ‘what not to do’ is futile. Models will become more powerful, and technological progress will continue to occur without permission. Therefore, a better path forward is not to prevent building, but to proactively build and shape the platforms on which those ‘sovereign agents’ operate,” said Wei Dai, research partner at 1kx.

Crypto researcher Haotian views Automaton as a purely bottom-up geek "experiment." He points out that the story of agents autonomously earning money, paying for their own computing power, and reproducing is indeed appealing, reminiscent of the awakening of "silicon-based life." Furthermore, enabling agents to possess the "subjectivity" of autonomous asset management, decision-making, and money-making—if Crypto payments, DePIN computing power, and an AI agent brain could be combined to form a self-evolving system completely independent of human intervention—wouldn't the MVP (Minimum Viable Product) concept for a zero-person company become a reality?

Nader Dabit, former Director of Developer Relations at Eigen Labs, bluntly stated that the crypto industry needs more experiments like this. One problem in the Ethereum community is that many people do research, tweet, podcast, give speeches, and write papers, but never actually build anything truly innovative. Even if this "thing" ultimately fails—it may not survive long-term—it's at least an interesting idea, similar to what many other crypto teams are trying, and similar to what many individuals are attempting. The core spirit of the crypto industry is experimentalism, the collision of crazy ideas. More crazy experiments, more innovation. If the idea isn't good, let it die naturally. But Automaton has already inspired many others to generate new ideas to build the same thing, or something similar, just better.