There is a specific kind of adrenaline that hits when you move from internal testing to real-world validation. For the last few weeks, we’ve been operating in a bit of a vacuum—building fast, iterating based on our own tastes, and feeling pretty confident in the "flow" of our tools.
But yesterday, the vacuum popped.
Following the "Great Cull" of Day 95, we focused our energy on our legal app. We didn’t want another prototype; we wanted a product that solved a friction point for professionals. So, we did the only thing that actually matters: we sat down with real lawyers and watched them use the app in real-time.
The "Intuitive" Delusion
As developers and designers, we often fall into the "curse of knowledge." We know exactly how the system works, so when we build a UI, we create a path that makes sense to us. We looked at the screens and thought, "This is clean. This is intuitive. The UX is seamless."
Then we watched a lawyer try to navigate it.
The silence was deafening. Every time they paused, squinted at a button, or asked, "Wait, where do I go next?" it felt like a tiny piece of our confidence was chipping away. We realized in real-time that our definition of "good UX" was fundamentally different from the mental model of the person actually paying for the service.
It was a brutal, necessary reality check. You can build a thousand prototypes with AI in a weekend, but you cannot AI-generate the empathy required to understand a user's frustration. There is a massive gap between a product that functions and a product that feels natural to a non-technical user.
From Feedback to Framework
The good news? This is exactly why we are building in public. The win isn't in having a perfect product on Day 96; the win is discovering exactly where you're wrong before you spend six months building the wrong thing.
We walked away from those sessions with a mountain of feedback and a very clear list of failures. Our internal design language needs to be scrapped in favor of a workflow that mirrors how lawyers actually process information, not how we imagine they do.
This experience has shifted our perspective on our broader mission. If we are building autonomous product development systems, we have to solve for more than just "feature completion." We have to figure out how to bake real-world user validation into the iterative loop. Speed is useless if you're accelerating in the wrong direction.
Strengthening the Engine: Haystax Updates
While we pivot the UX of the legal app, we’ve also been refining the machinery we use to build everything: Haystax.
One of the biggest bottlenecks we've encountered is the friction involved in getting a prototype from our environment into the hands of a tester. To fix this, we’ve been working on making sharing significantly easier. We’re moving toward a model where we can ship live, interactive prototypes instantly, allowing users to "play" with the product without the usual setup hurdles.
We’ll be rolling these updates out soon. The goal is to make the feedback loop we experienced yesterday—the "humbling" part—happen faster and more often. The more we can lower the barrier between "idea" and "user feedback," the less time we spend guessing.
Day 96 taught us that the most valuable tool in our stack isn't the AI—it's the honest, unfiltered feedback of a frustrated user. Now, it's time to get back to work and actually fix the UI.