AI learns to code faster on the Internet Computer, says Dominic Williams

Dominic Williams, founder and chief scientist of the Internet Computer Protocol (ICP), has highlighted a shift in how artificial intelligence is beginning to build apps on-chain. He said AI can now generate complex backends directly on the ICP blockchain using its Motoko programming language—several times faster and cheaper than on conventional infrastructure.

Williams explained that while ICP’s strengths include security, resilience, sovereignty and multi-chain connectivity, the real breakthrough lies in how AI interacts with the network. Instead of just using AI as a tool for automation, developers can now rely on it to write, deploy and manage code entirely within a decentralised environment.

The ICP blockchain allows AI models to run natively on-chain, removing the need for traditional cloud services. This design means the code an AI produces can execute directly as a smart contract without external dependencies. By using Motoko, a language purpose-built for ICP, the process becomes faster and more cost-efficient, lowering the barrier for developers and organisations exploring Web3-based applications.

Williams described this evolution as an early glimpse of a “self-writing internet” where developers describe what they want and the network builds it for them. It marks a potential shift in how digital products are conceived—away from manual coding towards systems that can understand intent and handle deployment autonomously.

The idea has sparked interest across the developer community, with some seeing it as a step toward democratising app creation. If AI can produce entire backends on-chain, individuals and small teams could build at the scale once limited to larger enterprises.

Others remain cautious, noting that blockchain-native AI still faces challenges around scalability, integration and energy efficiency. There are also questions about how much control developers retain once AI takes over complex coding tasks, and how auditing or security reviews will adapt to machine-generated systems.

Even so, the direction Williams describes suggests a broader movement within Web3—one where AI and blockchain are no longer treated as separate technologies but as complementary forces. If successful, ICP’s approach could redefine how quickly software is built and how ownership of that software is shared across decentralised networks.