The bug I would never have caught alone: my first week of beta
This blog is new, and I am opening it with a promise: I will document building Verbasil honestly, the wins and the misses. You learn more from the misses. So let us start with a miss that taught me a lot.
I tested my product as if I had not built it
Last week I stopped looking at Verbasil through a developer's eyes. I created a fresh account, like any new user would, and went through the whole journey from the start: sign up, first import, waiting for the memory to build, landing on the cockpit.
That mental switch is harder than it sounds. When you build a feature, you always test it slightly rigged: you know where to click, you have the right data, and without noticing you avoid the paths that break. You validate what you meant to build, not what the person in front of you will actually experience.
The moment it broke
On the cockpit, Verbasil is supposed to surface the decisions that need attention, with their evidence. And there, right in front of me, decisions appeared. Then they vanished.
No error message. No broken screen. Just real decisions, shown for a second, then gone on the next reload. The nastiest kind of bug: silent. If I had not been looking at the screen at the right moment, actually using the product, I would have missed it.
Why I would never have caught it alone
Here is what struck me. This bug did not trigger while coding. It did not trigger in my usual tests. It only appeared under real usage conditions: a real account, a real import, the memory building in the background while you watch.
In other words, I could have kept coding for weeks on my own and never run into it. That is exactly the trap of the founder who builds alone: you tell yourself a reassuring story in front of your screen, and you confuse "it works for me" with "it works."
I tracked the problem down to its root cause, in the machinery that builds the memory, and fixed it. But the real lesson is not technical.
The irony I did not miss
Verbasil exists for one reason: to make sure a company stops forgetting its decisions. And my first serious bug was exactly that: decisions disappearing.
The product meant to fight forgetting was forgetting. It is hard to find a more direct reminder of why this problem is hard, and why it deserves to be taken seriously.
If the underlying topic interests you, I wrote a dedicated piece: Why startups forget their decisions.
What I am changing from now on
One conviction came out of this week stronger than ever: nothing replaces putting the product in people's hands. The field does not lie. In front of your own screen, you tell yourself whatever you want. Faced with a real journey, the product tells the truth.
Concretely, that means:
- testing in real conditions, not developer conditions;
- letting people who are not me try the product live;
- treating every piece of field feedback as a gift, especially when it stings.
It is less comfortable than polishing a feature on your own. It is far more useful.
What is next
I am in the middle of the beta, in direct contact with the first people testing Verbasil, and I have set a clear target: earn my first euro very soon. I will write about that here too, whether it works or not.
If you are building something right now: get it off your screen. Hand it to someone. Watch what actually happens. It is uncomfortable, and that is exactly why it is worth it.