Designer Explains Why You Keep Switching AI Tools — and What GPT Voice and Apple's Breathe Get Right
15 hour ago / Read about 9 minute
Source:TechTimes

Connie Zhou entered AI not through ChatGPT but through multimodal music and audio tools. As a product designer behind Mathos AI (YC-backed EdTech) and Dora AI (No-Code Creative Tools), she builds tools that teach through use. Her method—the CLAP framework—boils down to one bet: people learn by making, not by being told. So, hand them the cupcake mix, and break some eggs.

Connie Zhou Connie Zhou | https://conniezhou.design


Irene: One of the things we forgot ChatGPT did was define what AI as a product looks like. Is there something about that interface you would change?

I wasn't an early ChatGPT adopter. I was already working with multimodal AI around 2020—audio models, generative music, motion graphics for WeJam, a guitar-learning platform. A text chatbot felt disconnected from everything I cared about. What pulled me in was OpenAI's audio capabilities and Google Creative Lab's music models. I didn't adopt AI because I was supposed to. I adopted it when it met me where I already was. Everyone does.

Irene: The magical moment that feels simple, intuitive, that you barely notice—it's the result of complicated engineering. Why does GPT's voice mode get that right?

Because it's a sensory shift, not a feature upgrade. Typing is 40 wpm—you're editing yourself before the thought lands. Talking is 140—just sentence dump. Voice handles nuance, vague preferences, even uncertainty. Good design means you don't think about it.

Connie Zhou | https://conniezhou.design

Irene: There's a dilemma—I want more people to know about this, but the whole point is you don't think about it. Is that why people don't know about design?

That's my pain point—and why I've been exploring a design producer role. "Designers don't brag." Chinese takeout boxes unfold into plates. Nobody knows. I want to be the translator of designers. As the magician LiuQian puts it: the magician isn't keeping the secret from you—they're keeping the secret for you.

Connie Zhou | https://conniezhou.design

Irene: Telling me what the Apple Watch haptic curve looks like means nothing. But telling me how you translate an emotion into a pattern—that's cool. What's the Breathe app doing?

The haptic vibration mimics human breath rhythm—elongated at the start and end to slow you down. The alarm is completely different, urgent. The meditation curve syncs with the flower animation. Haptic, visual, rhythm—harmonized like an orchestra. You feel "this is nice" without knowing why.

Connie Zhou | https://conniezhou.design

Irene: How do you get users to enjoy the process?

The Fogg Behavior Model: B = MAP. Everyone focuses on motivation. The overlooked part is ability and prompt—make it easy to start and worth staying in. I apply CLAP—Creation-Led Adaptive Pedagogy—across everything: start from what people care about, get them making immediately, engage multiple senses, let AI adjust like a tutor, embed teaching inside the tool.

Irene: For the rest of us, the mundane is just what you endure. It was inspiring to think of it as something to turn our attention towards.

That's the through-line. Life is the process—showering, working, creating, playing, thousands of times over. The AI tool you'll actually keep won't have the best benchmark. It'll be the one that makes the everyday worth staying in.


Connie Zhou is a product designer and founder of Meso Design. Her work spans EdTech and No-Code Creative Tool, and has received the iF, IDA, Core77, and A' Design Award.

Irene Zhang writes about AI and politics for ChinaTalk. Cited by the New York Times, Semafor, and RAND. Subscribe at www.chinatalk.media.