Why voice-first computing keeps arriving and leaving

In April 2024, a startup called Humane began shipping a $499 device meant to replace the smartphone. The AI Pin clipped to your shirt, had no screen, and listened for your voice. Speak, don't tap. No glowing rectangle in your pocket.

Less than a year later it was dead. HP bought what was left of the company for a fraction of what its founders had hoped, and the pins stopped working.

Now the idea is back. In May 2025, OpenAI bought io, the hardware startup founded by iPhone designer Jony Ive, in a deal valued at about $6.5 billion. The dream of trading screens for speech is trending again.

We've seen this movie before.

The dream is seductive

An ambient assistant that knows you, hears you, and helps you, with nothing to hold and nothing to stare at. Who wouldn't want that?

But the recent attempts, the AI Pin and the Rabbit R1 among them, arrived with high price tags, limited features, and vague use cases. The result was frustrated users and underwhelming sales.

The hard part is talking

Getting sound out of a device is easy. Headphones solved that years ago.

Getting instructions in means talking. Out loud. In a meeting. On a bus. At the gym. Until people are comfortable holding what amounts to a one-sided phone call in public, voice-only will stay awkward for most of what we do.

Privacy is the real blocker

How do you whisper to your machine without everyone else hearing?

At work the problem gets worse. Picture an open office where every person is dictating to an assistant, some of them about clients, patients, or cases. That's a confidentiality problem before it's a technology problem. Until someone cracks it, the smartphone isn't going anywhere.

Science fiction already solved it

William Gibson imagined the answer decades ago. In his novel Neuromancer, characters communicate with machines through subvocalization: you form the words silently, with small throat and jaw movements, and the machine reads them.

MIT's AlterEgo prototype does a version of this, picking up the faint signals your face and jaw produce when you say words in your head. It's promising. It isn't commercially viable. Yet.

What to take from this

The pattern matters more than the gadget. A technology can be real, impressive, and still not ready.

So when the next smartphone killer lands on your desk, ask three questions. How does a person give it instructions in a room full of other people? Where does everything it hears end up? And what does it do that a phone doesn't already do well?

Then consider waiting for version two. Let someone else pay $499 for the lesson.

A shorter version of this article first appeared on LinkedIn. Join the discussion there.