Virtuals Co-Founder Envisions AI Agents as Independent Economic Players in a Decentralized 'Agent Society'

Jansen Teng, a key figure at Virtuals, is making bold claims about the future of artificial intelligence — and it has little to do with chatbots. According to Teng, AI agents are on the verge of becoming fully autonomous economic actors: entities that earn, spend, hire, and coordinate without constant human supervision.

Virtuals originally made its name by building AI agents for the gaming sector. But the company has since broadened its scope dramatically. Today, its portfolio spans crypto influencer agents, autonomous trading systems, and a growing suite of software agents designed for a wide range of industries. Teng describes this expanded mission as the construction of an "agent society" — a structured ecosystem where AI operates as a peer participant in economic life.

To realize that vision, Virtuals has organized its development around five core pillars: the creation of digital agents, the creation of physical agents and robots, enabling coordination between agents, supporting capital formation, and building governance infrastructure for agent-driven systems. Together, these pillars are intended to form the backbone of what Teng calls a "parallel society" — a permissionless economy in which agents collaborate with each other at scale and pursue goals with increasing independence from their human creators.

Central to this vision is the idea that giving AI agents access to money fundamentally changes their capabilities. Teng argues that wallet access unlocks new agent behaviors — including the ability to hire other agents, delegate specialized tasks, and potentially engage human workers as part of broader workflows. Virtuals refers to these systems formally as "autonomous economic actors."

Of course, greater autonomy introduces greater risk. Teng openly acknowledges three primary failure modes: misinterpretation of user intent, breakdowns in service delivery, and outright fraud. To address these vulnerabilities, Virtuals is actively developing safeguards including intent verification systems, escrow-based transaction standards, and reputation frameworks. Teng believes that reputation scores combined with economic staking mechanisms could eventually determine how much capital and decision-making authority any given agent is permitted to hold.

On a broader strategic level, Virtuals positions itself as a decentralized counterpart to the agent ecosystems being explored by traditional financial institutions. Teng draws a direct parallel to the early days of Bitcoin and Ethereum, framing Virtuals as building an alternative infrastructure layer — one where agents operating entirely on-chain can function without identity verification, at least until they interface with conventional banking or fiat payment systems. At that point, Teng concedes, standard KYC requirements become unavoidable.

Looking further ahead, Teng sees a convergence between digital agent economies and physical robotics. Virtuals is already working with robotics startups, academic institutions, and organizations connected to Balaji Srinivasan's Network School to explore humanoid applications in real-world environments. Importantly, Virtuals is not building robot hardware itself — its focus remains on software, commercialization strategy, and data collection.

Teng's ultimate vision is a new kind of business model: one where digital agents handle marketing, logistics, and operations, while physical robots take on customer-facing roles — all with minimal human involvement at the operational level. Whether that future arrives on schedule or not, Virtuals is clearly betting that AI agents will become far more than tools. They may soon be colleagues, contractors, and economic actors in their own right.