Life’s a game, and AI agents are Animoca’s new play

Don’t call it a pivot.

Yat Siu, co-founder and executive chairman of Animoca Brands, spent eight years selling the Hong Kong-based company as a Web3 gaming platform. Now Animoca has launched an agentic AI platform called Minds, and Siu is ready to argue as soon as the often-contentious term is mentioned.

“It’s not so much of a pivot – it’s an evolution because we haven’t gone away from [gaming],” he says.

Yat Siu, co-founder and executive chairman of Animoca Brands / Digital collage by Ulla, photos courtesy of Yat Siu

The distinction matters to Siu, as it frames a harder question to which Animoca appears to have given a fair amount of thought: Does Minds mark a genuine change from blockchain-based gaming into AI, or is the company simply adding one of the hottest technologies in years to its core business?

Siu argues the latter, but the evidence cuts both ways.

A platform built for scale

Launched in February, Minds runs customizable agents in the cloud instead of on a user’s machine. Siu says the platform works over email, Telegram, and WhatsApp and doesn’t require dedicated hardware or code skills.

Minds pools together more than 50 AI models. MiniMax M3, a Chinese open-source model that launched in June, does most of the work, while Gemini, Grok, and Claude handle more specific tasks. According to Siu, most Minds users never even learn which model answers them, and the system optimizes for cost and output behind the scenes.

Animoca has launched Minds, an agentic AI platform, but Yat Siu says the company is still in gaming. / Photo credit: Animoca Brands

He explains that the platform is based around “skills.” The Minds marketplace holds roughly 5,000 of them, and building one rarely means writing code. A user can direct an agent toward a GitHub repository, ask it to turn the code into a skill, and publish the result within an hour. The first “builder” pays the token cost.

Everyone who loads the skill afterward pays a fraction, and the creator earns a cut.

See also: Firms are spinning off AI tools, but are clients interested?

Siu says he runs about 300 agents of his own and earns somewhere between US$600 and US$1,000 a month on the skills, which he considers analogous to smartphone apps.

The case for evolution

Can it pay for itself?

Where Web3 stops looking like baggage

Can agents win the revenue game?