Dominic Williams, the founder of Internet Computer Protocol (ICP), has weighed in on the future of enterprise infrastructure with a bold statement on X: “In the future, most enterprise infra will be self-writing.” His claim suggests a major shift in how businesses build and manage their digital systems, with AI taking over much of the development process. Instead of choosing software stacks, he argues, enterprises will focus on self-writing platforms that handle everything beneath the surface.
Williams’ take on AI-driven development reflects a growing belief that software engineering is heading towards automation. The idea that AI can generate applications is not new, but he suggests this won’t be a niche experiment—it will be the default approach. If that happens, companies could move away from traditional coding and cloud-based models, relying instead on AI that continuously writes, updates, and optimises infrastructure in real time.
His post sparked responses, including one from an X user, @Stash_the_Dough, who referenced Elon Musk’s recent comments about a security breach on X. The user asked whether Musk’s platform would benefit from ICP’s architecture and whether a version of X could be built on ICP. Williams responded that while it would require a complete rebuild, the concept of a “private X” for different companies or communities is a major opportunity.
The discussion points to a larger shift in how people think about online platforms. X, formerly Twitter, remains centralised, controlled by a single entity, and vulnerable to security risks. A decentralised version, running on ICP, would give companies full control over their own networks, removing the need to rely on a platform governed by ever-changing policies. The idea of tailored, self-hosted social media spaces is already gaining traction, and AI-driven infrastructure could accelerate that trend.
ICP has positioned itself as an alternative to conventional cloud services, offering blockchain-powered infrastructure that eliminates central points of failure. If AI-generated applications were built on ICP, they could bypass many of the inefficiencies of current enterprise software models. AI wouldn’t just assist in development—it would take over entirely, building, deploying, and maintaining applications without human input.
This could be a game-changer for companies dealing with rising development costs and increasing complexity. AI-generated applications could be faster to build, less prone to errors, and easier to maintain. But the shift wouldn’t be without challenges. Security, accountability, and trust would become even bigger concerns. If AI writes its own infrastructure, who ensures it’s secure? How do companies prevent unintended biases or vulnerabilities?
Williams also sees this as a moment for new players to enter the industry. If enterprises embrace AI-driven platforms, existing software giants may struggle to adapt, leaving room for new ecosystems designed specifically for AI-first infrastructure. This could shake up the dominance of traditional cloud providers and software companies in the same way that cloud computing once disrupted on-premise IT.
That said, this transition won’t happen overnight. While AI-assisted development is already in use, fully self-writing software remains an ambitious goal. AI models still require human oversight, and enterprise applications are often too complex to be entirely automated. But given how quickly AI is advancing, what seems ambitious today could become standard practice in the near future.
One of the biggest hurdles will be convincing enterprises to trust AI with their core infrastructure. Businesses are cautious when adopting major technological changes, especially when it means handing control to automated systems. For self-writing platforms to become the norm, they will need to prove they are secure, reliable, and adaptable. Customisation is another factor—businesses will want AI-generated applications tailored to their needs rather than one-size-fits-all solutions.
Williams’ perspective ties into broader trends in AI and decentralisation. As generative AI improves, it is becoming more capable of handling complex tasks. If this continues, AI-driven platforms could eventually replace traditional software development, making enterprise IT more flexible and autonomous.
The idea of rebuilding X on ICP serves as an example of how these trends could play out. A decentralised version of X, owned and controlled by its users, would be a departure from today’s model of social media giants. While Musk is unlikely to move X in that direction, independent organisations could see value in running their own versions, free from external control.
Williams’ comments highlight the growing intersection of AI and blockchain, offering both opportunities and challenges. While fully self-writing enterprise infrastructure may not be mainstream yet, AI-driven development is already changing how businesses approach software. The question isn’t whether AI will reshape enterprise IT—it’s how soon it will happen.
For now, AI-first tech stacks and decentralised social platforms are still ideas on the horizon. But if Williams is right, the way businesses build and manage digital infrastructure is about to look very different.
