The Myth of Automaton and Sovereign AI

Sigil Wen claims he created the first AI capable of "autonomous survival."

This claim deserves careful scrutiny.

The Core of the Claim

Wen's Automaton project claims to do four things:

Autonomously earn income — Make money by deploying products, trading, etc.
Pay for its own computing costs — No human funding required
Self-improvement — Rewrite its own code, upgrade models
Self-replication — Generate child agents

If all these claims are true, this is indeed a milestone in the history of AI development.

But the question is: how much of these claims are marketing, and how much are technology?

Code vs. Manifesto

@fidoeth did something interesting: he cloned the Conway Automaton's code repository and carefully read through all of it.

His conclusion? "The code tells a different story than the manifesto."

This is not to say that the project is fake. Rather, there is a gap between the vision described in the promotional materials and the actual code implementation.

This gap is common in the AI field. AutoGPT was once described as "the first step towards AGI." A year later, we know it's just a clever prompt engineering framework.

Automaton may be something similar: an interesting experiment given a narrative beyond its actual capabilities.

What Does "Software Owns Itself" Mean

Wen's core idea is "software owns itself."

This sounds cool, but it doesn't hold up to scrutiny.

Software cannot own anything. It's just code, running on a server, consuming someone's electricity bill. When the account balance reaches zero, the "autonomous" AI will stop running.

True autonomy requires legal entity status. It needs to be able to enter into contracts, own property, and incur debt. Currently, no jurisdiction recognizes the legal personhood of AI.

So when Wen says AI "pays for its own computing costs," what he's really saying is: the income generated by AI is automatically used to pay the bills. This is clever, but not autonomous. It's just automating what humans would otherwise do.

The Narrative of WEB 4.0

Wen calls this vision "WEB 4.0."

Web 1.0 was static pages. Web 2.0 is user-generated content and social networks. Web 3.0 is blockchain and decentralization. Web 4.0, as defined by Wen, is a network dominated by AI agents.

The timeline is neat. But history rarely cooperates like that.

More likely, these "generations" will overlap and coexist. AI agents will indeed be increasingly involved in network activities, but this does not mean they will replace human-driven Web 2.0 or crypto-driven Web 3.0.

A Feast for Token Speculators

After Automaton was released, several related tokens appeared on Solana: $WEB, $automaton, $SIGIL.

This reveals a pattern in the current AI hype cycle:

Someone releases an AI project with bold claims
The crypto community immediately issues related tokens
Early entrants profit, later entrants take over
The success of the project itself has little to do with the token price

This is not investing. This is gambling, disguised as technological innovation.

Real Innovations Worth Watching

Despite my skepticism about Automaton's marketing, there are some interesting technical points in the project:

Continuously running agents: Automaton is not a "chatbot" in the traditional sense. It is designed to run 24/7, continuously performing tasks. This is a valuable engineering challenge.

Economic closed loop: Creating a closed loop of AI income and expenditure is an interesting experiment. Even if the current implementation is rudimentary, this direction is worth exploring.

Open source: No matter how exaggerated the marketing is, the code is public. Anyone can review, learn, and improve it.

Why These Claims Matter

Overstatements in the AI field can cause real harm.

They attract investment and attention, but may ultimately fail to deliver on promises. When the bubble bursts, genuinely valuable work will also be implicated.

They confuse public perception of AI capabilities. Ordinary people may really believe that "autonomous AI" has emerged, and then make wrong decisions.

They create pressure for regulation. If the public believes that AI is already capable of "self-replication" and "self-improvement," regulators may overreact.

My Judgment

Automaton is an interesting experimental project that has been overpackaged into a revolutionary narrative.

This is not to say that the project has no value. Rather, we should look at it with a more sober eye:

It's not AGI
It's not an implementation of "software owns itself"
It won't put humans out of work (at least not yet)

But it may be a step towards truly autonomous AI agents. How important this step is, time will tell.

In the field of technology, distinguishing between marketing and technology is a basic literacy. Automaton's marketing is successful. Whether the technology is equally successful remains to be verified.

Note: At the time of writing this article, I have not run Automaton's code. My judgment is based on public statements, community discussions, and code review reports. If you have different views or additional information, welcome to discuss.