Not the obvious one, whether users like it, whether it grows, whether it works technically. But a quieter tension: whether what you’re scaling is merely useful, or whether it actually contributes to learning.
I’m still figuring this out.
The First Signal That Something Was Working
With SceneSnap, the first thing that clearly worked was our AI tutor.
It was different enough to feel original. People clicked with it. They came back. They told us it helped them move through material that had previously felt dense or boring. That was the first real signal that we weren’t just automating existing workflows, we were onto something new.
At that stage, disruption and usability were the main challenge. Make it powerful, but not overwhelming. Novel, but not confusing. We iterated constantly to find that balance.
And for a while, that felt like the core problem to solve.
When “Useful” Started to Feel Insufficient
Early versions of the product were undeniably useful. Automation alone created real value. Passive materials became interactive. Students could do more with less effort. That mattered.
But over time, another realization set in.
Usefulness wasn’t the ceiling. It was the floor.
We weren’t just building features that saved time. We were beginning to create a first digital learning process that couldn’t exist without AI, something that didn’t just assist learning, but started to feel necessary for certain types of understanding.
That shift changed the questions we had to ask ourselves.
Making Learning Easier Is Not a Single Thing
What we’re doing sits on two fronts.
On one side, we reduce unnecessary friction. Boring, passive materials become interactive. Learners can engage instead of endure. That’s clearly good.
On the other side, we try to preserve what makes learning effective in the first place. The best tutors don’t just explain. They adapt. They personalize. They notice where understanding breaks down and respond accordingly.
AI lets us approximate that at scale.
But this is where the tension appears.
The Tension I’m Living With as a Founder
As a founder, I feel caught between scaling something that is broadly useful and scaling something that genuinely supports cognitive development.
Those two things overlap, but they are not the same.
Making things easier can help learning. Making things too easy can quietly undermine it.
Removing friction can unlock engagement. Removing the wrong friction can flatten understanding.
This isn’t a problem you solve once. It’s a balance you keep renegotiating as the product grows.
What I’m Learning (Without Pretending It’s Settled)
I don’t think learning products fail because teams don’t care about learning. Most do. I think they fail because it’s hard to see the difference between comfort and progress, especially at scale.
We’re still iterating. Still watching how people actually use what we build. Still adjusting when something “works” but doesn’t lead where we hoped.
The hardest part isn’t making learning tools people enjoy using. It’s making tools that people enjoy using and that help them grow in ways that aren’t immediately visible.
That’s the problem I’m trying to solve.
And I don’t think pretending it’s already solved helps anyone.