When most people imagine the future of AI, they envision seamless automation, personal assistants with perfect recall, and intelligent systems that work tirelessly to anticipate their needs. What few consider is who will own those systems, who controls them, and what that means for the individual. In an age increasingly defined by rented intelligence—where corporations lease our tools back to us, gatekeep the best models, and monetize every input—David Minarsch, co-founder of Valory and founding member of the Olas Network, offers a radical counterpoint: the real AI revolution is about ownership.
This isn’t just a technical issue; it’s a philosophical and economic one. Minarsch’s work sits at the intersection of applied game theory, decentralized infrastructure, and agentic autonomy. His conviction is clear: in order to ensure AI serves people—not platforms—we must build systems that are owned, operated, and directed by users themselves.
From Closed Models to Open Intelligence
“Open source AI is this idea that I can like outright own the model… and host on the infrastructure of my choosing,” David explained during the CoinRock Show.
That simple premise upends the default assumption of today’s AI development, which is largely dependent on corporate APIs, black-box models, and centralized service layers. Instead of renting computational power from tech giants and relying on proprietary weights, open source AI would allow builders and users to run models locally, upgrade them collaboratively, and govern their direction.
This vision isn’t merely about cost savings or decentralization for its own sake. The deeper issue is that as businesses increasingly depend on AI for daily operations, the centralized rent-seeking model distorts incentives, limits competition, and erodes the sovereignty of users. As Minarsch put it,
“Some fraction of this [revenue] is starting to grow that is basically, you know, costs to OpenAI or Anthropic or whatever model you’re calling.”
By embedding AI models within tokenized, community-owned ecosystems, projects like Olas propose a counterweight to this dynamic. AI agents become tools you actually own—not rented intermediaries you must ask permission from. This is a fundamental inversion of power in the age of artificial intelligence.
Autonomy by Design, Not Subscription
Much of Minarsch’s work focuses on autonomous agents—not just large language models, but full software stacks that act on behalf of users in complex environments. These agents are programmable, self-directed, and capable of delivering value without human micromanagement. But perhaps the most compelling part of his argument isn’t technical. It’s personal.
“How do we get back to this world where we have autonomy for our software and autonomy for ourself,” he asks, invoking an era when ownership was tangible and absolute.
The devices we used didn’t come with remote kill switches or monthly billing cycles. They belonged to us. The AI-enabled future, by contrast, risks becoming one giant perpetual lease.
Consider the proliferation of SaaS-based AI tools and subscription-bound assistants. Even our smart TVs now stop working without corporate permission.
“We have these weird hybrids where even the TV, we’re sort of renting it because the TV will stop working once certain subscriptions are turned off,” Minarsch observes. The implications are chilling when applied to AI.
Ownership of AI agents isn’t just about ideology. It’s about resilience. Permissionless systems, like those built on crypto rails, allow agents to act without being throttled, de-platformed, or denied access due to arbitrary terms of service. In a future where AI is embedded in finance, governance, and communication, this kind of resilience could mean the difference between empowerment and dependence.
Tokens, Incentives, and Coordinated Intelligence
The Web3 stack offers a unique toolkit for enabling truly owned AI. Minarsch’s team at Valory is pioneering a structure where AI agents are both owned and improved by users, with token incentives tying these actors together into multi-agent networks. Unlike traditional products that extract value, this model distributes it.
“We’re using incentives to coordinate like a network of builders and operators and other kinds of contributors… to bootstrap an ecosystem of user-owned agents,” he said.
The idea isn’t just to decentralize infrastructure, but to decentralize development itself.
This notion of a “multi-agent economy” borrows from the natural world: different specialized actors, cooperating across a shared incentive structure. Much like humans, these agents will develop comparative advantages and specialize. They won’t all need to be generalist super-models. A DeFi agent, for example, won’t need to know how to write content or generate memes. What it will need is the ability to act, transact, and optimize on behalf of its user—without intervention.
“If you look at humans, they’re initially somewhat generalized… and then we specialize for economic and societal reasons. The same pressure applies to software agents.”
In such a world, tokens are not just speculative assets but coordination tools. They tie together contributors, developers, and users into ecosystems where ownership is spread and value is created collaboratively. It is in this space that crypto finally finds a compelling purpose beyond financial speculation.
The Future Is Agentic
What makes David Minarsch’s vision so compelling is not just its technical novelty but its ethical clarity. He is not building AI to make life slightly more convenient or to shave seconds off a task. He’s trying to restore a lost relationship between people and their tools. One where agency—true agency—means control over your systems, your data, and your outcomes.
“I would like a general purpose agent that I can give somewhat arbitrary goals and it will, within the means of its resources, execute against them. I would want that system to be super, super transparent and… I want to have ownership of it.”
The AI race is already underway. But the real question isn’t who builds the smartest model. It’s who owns the model—and whether the rest of us are just renting the future.
