Why voice will be the core interface

Training Data is an interview series where the venture firm Sequoia talks with people building AI. In this episode the guest is Mati Staniszewski, who co-founded and leads ElevenLabs, the voice and audio company.

The discussion is less about any single product and more about a direction: where voice technology is heading, why it is becoming central, and what it takes to build a durable company in a fast-moving field. Treat it as a window into how an operator thinks, not a technical tutorial.

The headline argument is that voice will be a core interface, not a side feature. For most of computing we have typed and tapped. As speech generation and understanding get good enough, talking to a device becomes natural and fast, and in many situations, hands-free or while moving, it is simply the better way to interact.

The second thread is focus. A common temptation, especially when a company is doing well, is to expand into many areas at once. The conversation argues for the opposite: pick the thing you are best at, here it is voice, and keep going deeper rather than wider. That discipline is presented as a feature of the strategy, not a limitation.

If voice really does become a main way people use software, then how you design products changes. You start thinking about conversations and sound, not only screens and buttons, and that reshapes what is worth building.

For anyone learning to build with AI, this kind of conversation is valuable in a different way than a paper. It teaches judgment: how a founder reads where the technology is going, how they decide what to chase, and why saying no to good options can be as important as saying yes. That is exactly the kind of thinking a guest lecture is meant to share with a community.