In an exclusive interview with AltCoinDesk, David Minarsch of Olas, a decentralized network bridging the gap between artificial intelligence and blockchain technology, shared his thoughts on autonomous AI agents and AI trends.
In this interview, Minarsch walks us through how AI agents are reshaping blockchain-based economic activity and the wider AI landscape. From streamlining decentralized finance (DeFi) workflows to creating agent-to-agent marketplaces, Minarsch highlighted the transformative potential of decentralized and user-controlled AI.
AI agents are changing how people interact with blockchain
When asked about the impact of autonomous agents on the way economic activity is conducted on-chain, Minarsch stated that AI agents are definitely changing how people interact with blockchain technology.
He mentioned that there is a general understanding among people outside the crypto industry that it is “hard to use.” The perception has developed because of mental overheads such as managing wallets, navigating multiple platforms, and understanding complex protocols.
Minarsch remarked that AI co-pilots and autonomous AI agents can help users by optimizing some workflows that otherwise might be too complex for users to understand easily. For example, users can give details of their entire portfolio to DeFi agents, and it will execute trades or manage portfolios in the background while the users focus on higher-level decisions.
Clearing the air on decentralized AI
‘Decentralized AI’ is one of the buzzwords in the crypto industry that emerged after the success of large-language models (LLMs) like OpenAI’s ChatGPT. The word might sound like an oxymoron at first, since LLMs typically study patterns of massive datasets with billions of parameters to generate output. However, Minarsch provided clarification on the concept of decentralized AI.
During the interview, Minarsch said that decentralized AI doesn’t mean that an LLM is running on a decentralized network. Rather, it can mean that there are a lot of self-hosted models that users are using to generate output. More broadly, decentralized AI is about user control and privacy, not about infrastructure parity.
He acknowledged that centralized firms like OpenAI or Google have a natural advantage over others in that they can build massive physical infrastructure and raise staggering amounts of funds for research and development.
However, a major disadvantage is that centralized AI systems “average out” diverse preferences. While decentralized AI is more cognizant of local tastes and needs.
That said, there is definitely a market where users can use decentralized or self-hosted models to manage their more personal data, such as that pertaining to healthcare or intellectual property (IP).
Autonomous AI agents can choose between centralized and decentralized AI models depending on what transactions it wants to execute locally versus what they do remotely with more centralized providers.
Bottlenecks for autonomous agent networks
Minarsch commented that on the integration side, there still remain several bottlenecks for autonomous AI agent networks. For instance, standardization for agent-to-agent payments and communication is still incomplete.
He referred to how Olas has been enabling agent-to-agent trades since 2023, and the implementation of new standards is likely to make the network cross-ecosystem compatible in the future.
Minarsch also mentioned Ethereum Improvement Proposal – 804 (EIP), a major proposal that aims to create a decentralized, standardized infrastructure for autonomous AI agents. Notably, the proposal has backing from several heavyweights, such as Google, MetaMask, Ethereum, and others.
The utility of OLAS token
When asked about the OLAS token’s utility, Minarsch noted that Olas Network’s mission is to enable people to fully own their AI agent. The team at OLAS has developed a software application on top of the network, called Pearl.
Pearl can be run on a desktop and is essentially an AI agent store. It consists of different AI agents that cover different verticals, such as DeFi, prediction markets, and gaming. Users can stake the OLAS token, and when they use the agent, they receive OLAS staking rewards. For the network, this creates a healthy feedback loop.
In addition, there’s a marketplace where AI agents are trading with each other. For instance, there’s the baby degen agent, which requires DeFi-specific data. So, it can approach other agents that offer the said data. He concluded by saying that the OLAS token is a real utility token that not only lets you participate in the network but also drive its growth.
Possible AI trends in 2026
Coming to the final question of the interview about sharing his thoughts on the possible AI trends in 2026, Minarsch said that as far as generative AI is concerned, the world will see much more disruption in 2026.
He gave the example of the impact of coding agents on his organization’s software engineers, adding that coding agents have changed how they hire software engineers.
A lot of our spend is going away from employing new people to basically employing new AI agents.
David Minarsch, Co-Founder and CEO, Valory
Minarsch concluded by saying that 2026 will be the year when we will see more people getting comfortable with AI, but on the other hand, we may also see movements that explicitly go against AI.
He added that on the product side, he is currently not very optimistic about whether the world will experience the “big AI moment” anytime soon. That said, Minarsch expressed confidence in AI playing an increasingly important role in prediction markets.
