With over 20 years of experience as a computer scientist, Dominic Williams, CEO and Chief Scientist of DFINITY, has been at the forefront of blockchain innovation. As the founding architect of the Internet Computer Protocol (ICP), he leads one of the most academically advanced teams in blockchain, which includes the highest number of PhDs working on any network.
In an interview with CCN.com, Williams shared how the Internet Computer is shifting blockchain narrative, moving from speculative assets to AI-powered, usable infrastructure. As of 26 March, ICP ranks 34th on CoinMarketCap with a market cap of approximately $3 billion.
During the interview, Williams explored a range of subjects, including the ‘self-writing’ internet, where anyone can build full-stack apps just by chatting with AI.
ICP’s Unique Approach: Running AI Models On-Chain
Williams began by highlighting a key technical milestone: ICP remains the only blockchain that runs AI models on-chain, not just controlling them via token wrappers or cloud APIs.
“There’s a lot of people talking about on-chain AIs, but what they mean is an AI running on Amazon Web Services with a token somewhere on-chain,” Williams said. “That’s not what we’re doing. We’re running actual AI models as smart contracts.” This capability required a rethinking of how blockchains work.
Williams explained that on ICP, AI models execute prompts in a decentralized environment without relying on centralized data centers. This makes them verifiable, censorship-resistant, and, most importantly, trustless.
Williams emphasized that this design lays the foundation for a new internet architecture where cloud providers don’t gatekeep. Running AI on AWS, or any centralized cloud provider, introduces gatekeeping risks by allowing third-party control over infrastructure that should ideally be decentralized and censorship-resistant.
How Self-Writing Apps Could Replace Traditional Development and Consulting
Williams made it clear that he views AI as not just an upgrade, but a total paradigm shift. “Self-writing is gonna completely change our relationship with tech,” he said.
Today, developers write code, deploy to servers, and maintain infrastructure. But tomorrow, according to Williams, users will simply describe what they want to an AI, and the Internet Computer will build and host the app, providing a secure URL for users to access it.
This vision is already being demonstrated through Caffeine, an internal DFINITY project that enables users to build applications in real-time through conversation. Whether it’s a private family photo-sharing app or enterprise-grade Customer Relationship Management (CRM) software, Williams explained that in around a year, users will be able to speak to the AI, and the AI will build the application.
“Retail users today have no chance of creating their own social media,” Williams said. “But this is going to be huge. Self-writing makes that possible.”
He added that self-writing will eliminate the need for companies to hire consultants, negotiate SaaS licenses, or even raise venture capital. “If you’re an entrepreneur, now you don’t need funding. You just talk to the AI.”
From AI-Assisted to AI-Native: The Future of Computing
Williams distinguishes between today’s AI-assisted development and tomorrow’s AI-native computing.
He introduced the term “vibe coding” to describe a process where AI helps developers write code, allowing developers to prompt, review, and deploy the results. With self-writing, however, that entire loop disappears.
It’s not just about helping you write code faster; It’s about skipping the coding step altogether,” Williams said.
In this new model, the AI doesn’t just assist—it writes, deploys, and updates software on its own. Williams believes this shift will compress the entire software development process into a simple chat interface, marking what he calls a “generational shift in who gets to build.”
Solving Data Loss Risks in Real-Time AI Coding: ICP’s Orthogonal Persistence
One of the biggest risks in real-time AI coding is data loss. Updating backend logic without breaking state is notoriously complex. Williams revealed that ICP has addressed this challenge through a system called orthogonal persistence.
“The logic and data of applications become one.”
Orthogonal persistence means that the state (data) is integrated within the logic (code), allowing applications to be updated without risking corruption or downtime. Williams explained that this breakthrough enables AI to write and safely evolve applications in production, even at “chat speed.”
The AI might hallucinate. That’s inevitable. But the platform must guarantee the data stays intact,” Williams explained.
ICP’s native architecture performs checks during migrations, ensuring that no referenced data is lost. This approach enables continuous improvement without the backend fragility that often plagues Web2 development.
How Chain Key Enables AI Apps to Interact With Multiple Blockchains
Beyond building apps from scratch, Williams envisions self-writing apps that interact with any blockchain using ICP’s chain key cryptography.
“The software doesn’t hold private keys. The blockchain itself signs the transaction.”
This innovation allows smart contracts on the Internet Computer to generate transactions on other blockchains without ever storing a private key. Instead, the protocol uses distributed key shares and continuous key resharing to ensure no one entity can control funds.
Williams sees this as a key enabler for a “world computer,” where all blockchains are unified through secure interoperability.
“Imagine you build a meme coin interface or a DeFi tool,” he said. “The app could live on the Internet Computer but talk directly to Solana or Bitcoin, using chain key behind the scenes.”
Power of AI to Reset Web3’s Focus From Tokens to Real Utility
Williams didn’t hold back when addressing what he sees as crypto’s lost potential.
“Crypto isn’t about zero-sum games. It’s about transformative technology that creates value,” he said.
Williams explained that tribalism, token speculation, and short-term hype have hijacked Web3’s original mission. “We’ve chased speculative gold, NFTs, memecoins. But this isn’t the work crypto set out to do.”
For Williams, the arrival of AI in the industry urgently calls for a reset. By enabling mass-market utility, self-writing apps could finally bring real users into Web3—not just investors.
Williams believes that “crypto has to be changed by force. I think AI is the force that will change crypto.” He went on to explain that “AI’s gonna clean the room. It’s gonna completely change how we think about what Web3 is.”
“AI is the forcing function,” he said. “The user doesn’t care what blockchain is under the hood. They just want a tool that works.”
Williams believes this utility-driven model will drive adoption far beyond today’s small pool of crypto developers. “Right now there are 25,000 crypto developers. But there are 5 billion people with smartphones. That’s the real audience, and we’re giving them superpowers.”
Was this Article helpful?
