You know that feeling when you open an app, get exactly what you asked for in 0.8 seconds, and then… nothing. No thrill. No curiosity. Just a hollow sense of “okay, done.”
That’s not a bug. That’s the design.
I spent a day at WAIC 2026 watching people interact with the latest AI products. And I saw something strange. Some products had lines of people genuinely excited, leaning in, smiling. Others had people who looked like they were waiting for a bus — bored, impatient, already scrolling elsewhere.
What was the difference? It wasn’t speed. It wasn’t accuracy. It was uncertainty.
At the NetEase booth, a user sat down and said: “Connect to my local project and run the test scripts.” The AI started moving the cursor, opening files, executing commands. The user watched passively. I asked him what he was waiting for. He said: “For it to finish.” He wasn’t exploring. He was auditing.
Across the hall, at the Qiyuan booth, two robots had no fixed schedule. A visitor walked in, not knowing what would happen. When she came out, she kept looking back. I asked why. She said: “I want to see what else it can do.” That “want to see” is curiosity in action.
Most people at WAIC chose the first experience. The crowd gathered around flawless demos. When a robot fumbled — a slight miss in grasping — the crowd dispersed. One engineer told me: “It grabs perfectly now because the lighting and position are fixed. In the real world, conditions change.” The audience wanted perfect, not real. Perfect means no surprises. No surprises means no curiosity.
Here’s the neuroscience bombshell: Your dopamine reward system doesn’t peak when you get the answer. It peaks in the 150-300 milliseconds before the answer, when you anticipate it.
That brief window of “I’m about to know” is where the magic happens. Too long, and anxiety kills it. Too short, and your brain never enters the anticipation state — the reward arrives before you wanted it.
Modern AI products are designed to compress that window to zero. “Book me a flight.” 0.8 seconds later, the result is on your screen. You haven’t even started wanting to know, and you already know. That’s not efficiency. That’s curiosity window closure.
Now, closing the curiosity window isn’t always bad. When you’re coding and need to know a function signature, you want zero friction. When you’re rushing to catch a flight, “the usual” is perfect. These are efficiency scenarios: clear goal, known path, verifiable result.
But when you’re exploring, creating, socializing, or discovering new content — these are experience scenarios. The value is in the process, not the destination. And here’s the trap: efficiency products are invading experience scenarios. Your phone’s AI writes a poem in 0.8 seconds. Your social app recommends people you “might like.” Your creative tool generates a finished piece from a prompt.
Users get speed. They lose the process. And they don’t even realize they’re losing anything, because there’s no toggle for “protect my curiosity.”
Based on what I saw at WAIC, I drew a line. Some products naturally protect curiosity: open-ended exploration, serendipitous discovery, non-deterministic outcomes. Others optimize it away: instant answers, predictive recommendations, one-click repeats. The problem is that the right column is infecting the left column, because “faster” is easy to measure, but “more curious” is not. Product managers have KPIs for DAU and task completion rate. They don’t have a KPI for “is the user still curious?”
The real issue isn’t that we close curiosity windows. It’s that we have no switch to open them when we want.
Imagine an AI agent that, when you say “book a flight,” responds in 0.8 seconds. But when you say “recommend weekend activities,” it pauses for 3 seconds, showing “Discovering things you might actually enjoy…” — not because it’s slow, but because the delay is designed. Or imagine a simple toggle: “Curiosity Mode.” Off: efficiency. On: exploration.
But here’s the deeper problem: users don’t know they want this. They don’t say “my dopamine prediction window is being compressed.” They say “this app is boring” or “I’m tired of this.” Product teams hear “boring” and add more features, better recommendations, more rewards. They’re solving the wrong problem.
The root cause isn’t lack of features. It’s the closure of the curiosity window. And the fix isn’t a smarter algorithm. It’s transparent window control — making the user aware that they are closing or opening their own curiosity, and giving them a choice.
So next time you feel that hollow satisfaction after an AI does exactly what you asked, ask yourself: Did I just get a result, or did I lose a moment of wonder?
Efficiency is a tool. Curiosity is a muscle. Don’t let the tool atrophy the muscle.
FAQ
Q: Isn't making AI faster always better?
A: No. Speed is only valuable in efficiency scenarios (repetitive tasks, known goals). In experience scenarios (exploration, creativity, serendipity), speed kills curiosity. The brain needs a brief delay to build anticipation. Without it, you get the result but not the satisfaction.
Q: How can I apply this to my product?
A: Audit your user flows. Identify where users are exploring vs. executing. Add intentional delays (2-3 seconds) with meaningful visual feedback in exploration flows. Consider a 'Curiosity Mode' toggle. Measure 'time spent wondering' not just 'time to completion.'
Q: Doesn't adding friction hurt retention?
A: Not if it's the right kind of friction. The goal is not to frustrate, but to create a predictable anticipation window. Think of it like a slot machine: the pause before the result is what hooks you. Games have used this for decades. AI products can too.