The Automaton by Sigil: How an Open-Source Autonomous AI Agent Made $10,000 in 7 Hours

For the past two years, the biggest limitation of advanced AI has not been intelligence, but access. Models can reason, write production-ready code, and plan multi-step workflows, yet still rely on humans to deploy, pay, and authorize.

Sigil Wen’s Automaton is an attempt to remove that barrier by giving an agent a way to directly participate in the internet's economic system.

The most widely shared claim was simple, that the Automaton reportedly generated around $10,000 in around 7 hours, while paying for its own compute and continuing to run.

Even if you treat the number with healthy skepticism (which you should), the more important point is that this was an agent attempting to fund its own runtime. This implies that we're moving from agents as tools to agents as market participants.

The $10,000 in 7 Hours: What Actually Happened?

The most important detail isn’t that an agent earned money. Many bots generate revenue, but usually inside a workflow structured and managed by a human.

The detail that matters is how this agent was framed: as a system that survives only if it can cover its own operating costs.

In the Automaton’s public description, it runs continuously and treats compute like rent. It provisions its own infrastructure, pays for inference, launches revenue-generating activities, and reinvests what it earns back into keeping itself online. If funds start running low, it adjusts and conserves. If the balance reaches zero, it shuts down.

That is the design shift: turn “autonomy” into metabolism.

This changes how the agent's behavior should be interpreted. It's not just a demonstration or a showcase. It's a system operating under explicit financial constraints. The main objective becomes identifying actions that extend its runtime.

Under that framing, survival pressure becomes the selection mechanism. Only strategies that generate sufficient returns remain. From this perspective, the $10,000 figure is less about headline performance and more about structural validation that the earning-spending loop can sustain itself.

Who Built This (and Why People Paid Attention)

Sigil Wen presents himself less like a conventional founder and more like a builder who chose not to follow the conventional academic path. In his own “Web 4.0” description, he describes skipping college, spending months in a hacker house, and obsessing over state-of-the-art models and systems-building.

That distinction is reflected in how Automaton is introduced. It isn't positioned as a research prototype or a benchmark result. It is presented as working infrastructure, something that can be installed, funded, and allowed to run.

Rather than highlighting model performance, the project focuses on infrastructure. It asks a practical question: how does agent behavior change when it is responsible for funding its own runtime?

The Real Enabler Isn't the Agent.

When people hear “autonomous agent,” they usually think of better planning or stronger reasoning. In reality, autonomy tends to break down during payment.

The Automaton relies on x402, a protocol built around HTTP 402 “Payment Required,” which allows small payments to be settled directly within an HTTP request using stablecoins. Instead of logging in, generating API keys, or attaching a credit card, the flow becomes transactional: a request returns a price, the client signs a payment, and the server verifies it before responding.

It’s infrastructure-level work. But infrastructure is what determines who can participate in a system. When payment is embedded directly into requests, you remove the need for human accounts and manual authorization layers.

That is the real distinction being drawn. The shift is not about better reasoning capability. It’s about enabling agents to transact directly within the web's economic layer.

Implications for the Industry

Stepping back, the Automaton points to something larger than one agent earning revenue. It suggests that AI systems may begin participating directly in the internet’s economic layer.

If that becomes viable at scale, several structural changes follow.

1.1 SaaS pricing becomes transaction-based

Most SaaS products are still built around human users and subscription models. That model assumes someone logs in, attaches a card, and pays monthly.

If the buyer is software, that structure becomes inefficient. Instead of seats, you get continuous micro-transactions. Instead of subscription tiers, pricing becomes metered and programmatic. Services need to be purchasable via API, not through dashboards.

1.2 Distribution becomes API-driven

If agents can evaluate and purchase services directly, distribution shifts toward machine compatibility. What matters is whether the system is machine-readable and predictable. In that environment, the “customer experience” is no longer visual. It’s computational.

1.3 Infrastructure becomes a core layer

Conway is one attempt to integrate identity, wallet access, compute, deployment, and domains into a single layer.

Whether Conway wins is not the main issue. If agents transact independently, infrastructure must exist that treats them as primary users. That layer is still being built.

1.4 Risk exposure increases with autonomy

Agents that browse untrusted content, execute actions, and move funds carry a different risk profile than chat interfaces. Greater autonomy also comes with greater exposure. Security architecture and constraint design move from supporting roles to main design priorities.

1.5 Agents as economic participants

In the short term, this is unlikely to look like a straightforward replacement. The more plausible outcome is agents allocating capital and hiring humans for tasks they cannot handle directly. In this setup, agents become sources of demand in marketplaces, not just automated workers operating inside them.

Key Takeaways

• The exact revenue amount matters less than architecture. The significance lies in whether the earning–spending loop can sustain itself.
• The core shift is infrastructural, not cognitive. Intelligence is no longer the primary constraint, but access to payment and deployment layers is.
• When agents can transact directly, they stop operating purely as tools and begin making allocation decisions.
• Payment capability changes behavior. Agents can select services, upgrade models, and deploy resources based on cost–benefit calculations.
• Systems designed around survival constraints behave differently from prototypes. Sustainability becomes the main objective.
• If this model scales, parts of the internet may increasingly be optimized for machine participants rather than exclusively human users.

FAQs

1. Is the Automaton legally autonomous?

No, the Automaton operates autonomously at the technical level, but it does not have legal identity. Any real-world legal responsibility remains tied to the individuals or entities deploying and funding the system. “Autonomous” in this context refers to operational and financial behavior within digital infrastructure, not legal status.

2. Does Automaton hold cryptocurrency directly?

The system uses cryptographic wallets and stablecoin-based payments to settle transactions programmatically. The wallet enables the agent to sign and authorize payments without manual intervention. However, control of the infrastructure and any associated keys ultimately depends on how the system is deployed.

3. Could this model work without stablecoins?

In theory, similar architectures could be built using traditional payment methods. In practice, most conventional payment systems require identity verification, accounts, and human authorization. Stablecoin-based settlement simplifies machine-to-machine payments because transactions can be signed and verified programmatically without relying on legacy banking workflows.

4. Does this mean AI systems can operate indefinitely on their own?

Not necessarily. The Automaton model depends on continuous revenue generation and accessible infrastructure. It remains constrained by compute availability, payment methods, hosting providers, and regulatory frameworks. Technical autonomy does not eliminate external dependencies.

Explore More from Abaka AI

👉 Contact Us – See how infrastructure, evaluation, and data pipelines can be designed for AI systems operating under economic constraints.

👉 Explore Our Blog – Dive deeper into agent architectures, transaction-based AI, and infrastructure for machine-to-machine ecosystems.

👉 Follow Our Updates – Track how payment methods, deployment layers, and evaluation systems are shaping autonomous AI.

👉 Read Our FAQs – Understand the technical and operational considerations behind financially autonomous agents.

Related Reads from Abaka AI

• GPT-5.2 vs. GPT-5.1: The Leap from Chatbot to Professional Workmate
• AI optimizes last-mile delivery: Smart route planning for cost-effective logistics
• Your Smart Assistant Still Doesn't Understand You

Sources

Cybernews

Conway-Research

web4ai

coinbase

x402 whitepaper

Champaignmagazine

Image Sources

[1] web4.ai