AI Chat Only Became Powerful After Structure Existed
The AI assistant sounded intelligent from the start. But it only became truly useful after the product had enough structure underneath it.
A curated collection of reflections from building ToGoStory - covering product judgement, design clarity, engineering realities, and leadership lessons along the way.
I started by consuming APIs, not building AI. What surprised me was that the intelligence only became meaningful after the foundations already worked.
I started by trying to animate everything. I finished by deciding what should stay still.
It wasn’t a big release. But it changed how the product feels.
A small feature turned into a lesson about control, simplicity, and judgement.
What looked like a small enhancement became a lesson in context.
The obvious choice turned out to come with quiet constraints.
What looked like a simple integration turned into a lesson in trade-offs.
Things were moving fast. Faster than I was comfortable with.
The more tools I explored, the harder the decisions became. Not because there weren’t options — but because there were too many.
The interface looked fine. But it was asking me to think too much.
We had everything built. And yet, something still felt wrong.
Everything was working. Until I realised I couldn’t trust what I was seeing.
I thought I was just trying something small. I didn’t expect it to change how I think about building.