We are so used to do-it-all tools that cater to so many edge cases, but I think this will change soon if AI-aided coding becomes the norm. We might see the trend shift to the utility approach and in turn, disposable software.

In 2026, if the goal is to use Ai to code, we should aim for smaller tools or utilities, that does just one thing with very little complexity. Such tools can be developed and deployed on a whimsy if you have enough tokens to burn.

This brings me to thoughts around disposable software that can be prompted into existence from scratch. These Ai generated software could be free of technical debt as it needs to meet the needs of today and just that. This will not make its way into enterprises, but I think we will see more of it in small to medium companies.

So instead of regular maintenance we could hypothetically create something new with an updated framework or stack while shedding the baggage of the past.

Such a shift would require a bigger change in how we think of software development and maintenance. The bugs that get prioritised for development would just be test items for the new development. Instead of incremental code changes and refactoring the previously developed, and sometimes flawed logic, we will make it do what we want today.

This approach still relies heavily on proper business documentation of what is being built and why; more business requirements and less of the functional requirements. It also works best in use cases that do not need to tap into any defined organisational data, as that could make it less of a utility and more of a full-fledged application.

This is still a rough thought and I think I will revisit this later in the year.