The Age of AI Chaos, the Foundation of Software's Next Decade
He thinks we are living through a remarkable period in software. Every week brings a new AI model, a new coding tool, another way for people to delegate work to machines. Companies are pouring hundreds of billions of dollars into infrastructure, while an ordinary developer can build a product in just a few days. Things move so quickly that what feels groundbreaking today may become ordinary in a matter of months
Yet this chaos is more than a technological arms race. It is shaping the foundations of software for the next decade, much as the internet did in the 1990s and smartphones did between 2007 and 2010. Thousands of ideas are being tested. Most will fade away, but a handful may become standards the whole world relies on
The biggest shift may be in how people interact with software. For the past two decades, we have built applications that people must learn to use. To transfer money, you open a banking app. To book a ticket, you visit a booking site. In the future, software may be the one learning to understand and carry out what people want, rather than the other way around
AI would then do more than help us write code faster. It would change how software is designed and how it creates value. Ideas like agents, MCP, skills, and long-term memory may become as familiar as databases and operating systems are today. Businesses could have digital workers that understand their processes, draw on shared knowledge, and work together to get things done
A paradox is beginning to emerge. As writing software becomes cheaper, code itself may no longer be the competitive advantage it once was. Value may shift toward organizing knowledge, earning trust, and solving problems that truly matter to people
He does not know what software will look like in 2035. Many of today’s tools will disappear, and some enormous investments may never pay off. Yet these experiments and failures are quietly shaping a new order
This decade, we are learning how to make AI work. Next decade, we may build software on the assumption that AI already knows how to work
And perhaps the greatest software products of the next decade are already taking shape in the middle of all this chaos
❤️ cowriter aethery