World’s Smartest Podcast App

Kevin Smith is the founder of Snipd, an AI-powered podcast app built in Zurich for people who listen to podcasts to learn. The app's core insight: podcasts have become the world's largest knowledge library, yet most people forget everything they hear. Kevin didn't accept that as inevitable - he built an AI system around the natural moments of discovery that happen during listening, turning a passive medium into an active knowledge capture tool.

This conversation covers why Zurich gives founders an unfair talent advantage, how AI should be integrated natively into user behavior rather than tacked on top of it, how crowdsourced snipping data creates an entirely new discovery layer for podcasts, and why 74% of podcast listeners - a far larger market than most assume - are primarily there to learn. Kevin also shares the three podcast episodes that changed his life, and why Switzerland's culture of self-responsibility produces better products and cleaner recycling bins.

The Problem: You Forget Everything You Hear

Kevin's founding insight came from his own frustration. Commuting by bike in Zurich, he'd hear a Peter Thiel quote or a Grammarly founder share an exact insight he needed - and the moment he tried to do something with it, the friction killed the knowledge. Stop the bike, take out the phone, open the notes app, type it down, pocket the phone, start riding again. Do that three times and you stop doing it entirely. By the time you want to share it with a colleague, you've lost both the quote and the context.

The traditional podcast player is a repurposed music player - play, pause, skip. It treats a two-hour conversation between world-class operators as roughly equivalent to a song. Snipd's premise is that the right frame isn't a music player. It's a knowledge interface - designed around the insight-dense moments that make podcasts valuable, and the natural human behaviors that happen when you want to capture and share those moments.

AI-Native vs. AI-Added: Bringing AI to the User

Kevin draws a sharp line between AI-native products and products with AI added. The lazy version of AI integration puts a chatbot field somewhere on the screen and calls it AI-powered. The native version starts with a user problem, not a technology capability, and asks: how can AI eliminate the friction that exists in this exact moment?

For Snipd, that moment is on a bike, on a run, driving - hands occupied, something remarkable just said, 10 seconds before the next insight arrives. The solution isn't to make users stop what they're doing and interact with a text interface. It's a triple-tap on AirPods. The AI does the rest: locates exactly what was just said, generates a structured summary, saves the clip and transcript to your personal knowledge library. The user never broke their flow.

This is the principle: don't force the user to change their behavior to accommodate AI. Find where the user already is, identify the friction in that moment, and remove it with AI. The product behavior emerges from user behavior, not from AI capability searching for a use case.

Crowdsourced Insight: When Listening Behavior Becomes Discovery Infrastructure

A Snipd user once told Kevin: "Do you realize what you've built?" He wasn't talking about the podcast player. He was talking about the aggregate signal. When millions of people independently tap their headphones at the moments that hit them hardest, the frequency of snipping per episode becomes a crowdsourced quality signal unlike anything that existed before. An episode with 400 snips has demonstrably more insight density than one with 4. You can see it before you start listening.

This solves a problem podcasters have complained about for years: there is no good discovery layer for podcasts. The search infrastructure is poor, the categories are coarse, and the recommendations are dominated by show-level popularity rather than episode-level quality. Snipd's snip density data is episode-level and content-quality-driven - derived from actual listener engagement with specific moments, not aggregate download counts or celebrity host name recognition.

The shareability dimension compounds this. A two-hour episode is a hard ask for a colleague. A 90-second snip with a summary - the exact moment that made you stop in your tracks - is a lightweight, high-signal unit that gets consumed and often pulls the recipient into the full episode. The snip is the word-of-mouth vector, and every share creates a new node in the discovery network.

Zurich as an Unfair Talent Advantage

ETH Zurich has pulled decades of world-class technical talent into a relatively small city. Google's largest tech hub outside the US is there. Apple, Meta's Reality Labs, OpenAI, and DeepMind all have Zurich offices. The talent pool is dense and the competition for it - unlike in San Francisco - is manageable. You're not competing against 400 other AI startups for the same senior ML engineer.

Beyond talent, Kevin points to a cultural advantage: Switzerland's concept of Eigenverantwortung - self-responsibility. The social norm that each person is accountable for doing the right thing without needing rules to enforce it. That culture produces liberal, high-trust environments where quality compounds over time because the market rewards it rather than defaulting to the cheapest acceptable option. For a founder building a premium product in a world of free alternatives, that's not an abstract cultural observation - it's a go-to-market reality.