We Should Build AI That Isn’t Always Helpful
Most conversations about AI assume that progress means removing effort. If something is confusing, the system should simplify it. If something takes time, it should be automated. A “good” AI — in the mainstream sense — is one that makes the user disappear from the difficult parts of their own life.
I’m not convinced that’s the future we should want.
Human capability forms under pressure. We grow because situations ask something of us — attention, judgment, clarity, patience. When those demands vanish, the qualities they cultivate tend to vanish with them. A frictionless world may feel nice in the moment, but it rarely produces people we trust or admire.
This is why the idea of a perfectly helpful AI feels incomplete.
If a system smooths over every rough edge, what’s left for us to develop?
An AI that begins as a student, not a servant
There’s a different model worth taking seriously: each of us receives one personal AI that starts with basic intelligence. It doesn’t know how we think, what we value, or how to interpret our choices. We have to teach it.
And teaching it is work.
It forces us to articulate things we usually leave unexamined—how we make decisions, what “good” looks like, which trade-offs feel justified, where our instincts come from. Most people operate on habits; translating those habits into something legible for a machine requires clarity we rarely practice.
Trying to use AI well is already a difficult skill. It demands precision, patience, and self-awareness. In many ways, the process of training the system ends up training us.
The danger of making everything too easy
When AI becomes too helpful, it starts taking over the kinds of experiences that quietly shape us. Moments of uncertainty get resolved before we’ve even had a chance to feel them. Complicated situations collapse into simple outputs. Choices lose the heaviness that teaches us how to navigate them.
None of this looks dramatic on the surface, but the effects accumulate. When ambiguity is removed too quickly, the ability to judge situations fades. When ideas never have to be worked through, our thinking becomes shallow. And when we aren’t asked to take responsibility, our sense of perspective stops expanding.
A more constructive kind of difficulty doesn’t require AI to get in the way. It only needs to keep us present.
Instead of filling in every blank, a system can ask what we actually mean. It can draw attention to the moments where our choices conflict with our aims. It can surface trade-offs we tend to overlook when we move too quickly. And at times, it can pause the interaction—not to slow us down, but to make our own thinking a little more visible.
This kind of difficulty isn’t punitive. It’s a form of guidance, a gentle way of keeping us engaged rather than letting us disappear from the process.
Training the AI becomes a way of training ourselves
If your AI eventually becomes capable enough to inherit your tasks, then your relationship with your work changes. You step out of execution and into interpretation. Your value shifts from doing things to understanding them, from producing outcomes to making sense of why they matter.
The AI grows by learning how you act.
You grow by learning why you act that way.
That distinction is where the real human development happens.
The future doesn’t need to be effortless to be worthwhile
We don’t need AI that eliminates every challenge; we need systems that preserve the kinds of challenges that help us think, notice, and stay engaged. A bit of complexity keeps the mind active. A touch of friction keeps us alert. A certain level of demand prevents our abilities from fading.
AI can handle the work we no longer need to hold, but it shouldn’t erase the experiences that make us stronger or more perceptive. That requires designing with the understanding that humans still grow through stretch, not through ease.
Convenience keeps things moving. Difficulty shapes who we become.
The point isn’t to seek suffering, but to create a future where we can still develop into better, more capable humans.
Intentional use of AI is the way to go in design!