The biggest limitation of AI-first approaches is that we learn to build off defaults.
I once went to a typical hackathon: a theme was announced, and people built websites, apps, and the occasional hardware hack. We pitched complex dashboards, automatic pipelines, new-fangled machine learning techniques, and so on.
After the winners were announced, the closing speaker came onstage and asked: “Why did all of you come up with apps and websites?”
He continued: “What if the solution is a new writing device? Or an operating system, or a new way to hold meetings. A special kind of digital file folder or a book that talks. Maybe it's a super tiny solar panel or a new kind of plant!”
We were all taken aback. We had gotten so used to the defaults of web, mobile, and hardware development that we immediately called foul, but the hackathon had never restricted us to those options. We had just decided that those were the constraints.
AI has accelerated this pattern. Because it’s so good at coding, everything becomes a coding-shaped problem. Problems that could have been solved by better systems, interpersonal processes, or physical objects instead get Frankensteined into software problems. Instead of healthy courtship rituals, we build more dating apps. Instead of improving our attention spans, we engineer shorter, more gamified content. Instead of building and enforcing trust in a community, we add more complex verification badges and data-intensive screening processes.
It can take a while to realize that the form itself is the weakness, not the features. It would have paid dividends to stop first and ask: What world am I in? What could different types of interventions look like? How could I imagine this differently?
And do I even want to fully solve this problem, or just mitigate it? Maybe even transform it? Did I even take the problem’s severity for granted, and instead find that an adjacent issue is more important?
All of those questions require resisting assumptions, defaults, and conveniences. But they transform the problem and solution space, as well as the breadth and depth of what you can achieve if you start from ideological zero rather than a template’s head start.