My first talking computer was a Commodore 64. It ran a program called SAM, Software Automatic Mouth, and it sounded like a robot gargling marbles. I loved it anyway. We watched films like Electric Dreams, where a lonely computer falls for the girl next door, and we believed talking machines were just around the corner. That was the mid-eighties. The corner took a while.
A few weeks ago I had a conversation with a machine and forgot, for a sentence or two, that it was one. I was testing Sesame’s voice preview, the one everyone compares to the film Her, and the comparison is fair. It paused to think. It breathed. When I cut in, it stopped, listened, and picked my thread back up. Four decades after SAM, a synthetic voice had finally stopped sounding like a machine.
That moment is bigger than a clever demo. Voice is about to change how we deal with everything that has a screen.
Siri and Alexa were the wrong shape
We have had talking computers for over a decade. They were clunky, and the reason was structural. Siri and Alexa were turn-based: you spoke, it waited, it guessed where you stopped, it answered, usually badly. Under the hood the parts were brittle and the seams showed.
The stack has caught up, fast. Speech-to-text is near-instant now, and it works across languages. Text-to-speech crossed from robotic to studio-grade, and with Sesame into something that actually carries emotion. And the whole loop went full-duplex: it answers in a few hundred milliseconds and lets you interrupt mid-sentence, the way people actually talk. Sesame set out to cross the uncanny valley of voice, and they got most of the way there.
A voice that pauses to think, breathes, and lets you interrupt is a different relationship with the machine.
Forty years of workarounds
Almost every interface we have used for forty years exists because the machine could not understand us. Keyboards, menus, apps, the screen itself: workarounds for a computer that could not hold a conversation. Take that limit away and the workarounds start to look temporary. OpenAI and Jony Ive are building a screenless device for exactly this, due late this year. The bet is that the most natural interface, the one toddlers and grandparents already share, finally works. Voice makes the keyboard optional.
We are already building with it
All of it is in production.
In Craise, we use a voice agent to let founders pitch to an AI partner before they ever face a real investor. It asks, it pushes back, it follows up. It is the closest thing we have built to an actual in-person session, and founders treat it like one, nerves and all. And that is still scratching the surface.
In Sassy, our work platform, we went further and made voice the primary interface for HR. You talk to it the way you would talk to a colleague who happens to know every policy, every leave balance, every form. There are no menus to learn. You just ask. Both work for the same reason: below a certain delay, a tool stops feeling like a tool.
And then there is Antone, the one that earned this essay its title. Antone is a chief of staff who lives in your WhatsApp. You talk to him the way you would talk to a person who happens to run your day: your mail, your calendar, the reminders and errands and small logistics that quietly eat a life. He has a voice, a memory of what matters to you, and a point of view. He is the most literal answer we have to the question the film asked. He is in daily use, he has a name, and he shipped.
The friend who always answers
A voice that listens without judgment, remembers you, and never tires of you is, for a lot of people, a real gift. If you are shy, or socially anxious, or grieving, or elderly, or just awake at three in the morning with nobody to call, a patient voice that always answers is a lot. I do not think we say that out loud enough.
A companion that is always warm, always available, and never difficult slowly raises the price of real people, who are messy and effortful and sometimes unkind. The thing that helps you cope can, if you let it, help you disappear. The comfort is real. So is the trap. And the cruel bit is that the exact qualities that make it a good companion are the ones that make it easy to lean on too hard.
A companion that never tires of you is a comfort. It can also quietly raise the price of every human who does.
There is no honest "good" or "bad" here. Whoever builds these things carries the weight of that trade, as part of the build and not a footnote.
That is the line we drew with Antone. He is warm, but he is not pretending to be your friend. His job is to take the logistics off your plate and hand you back the hours, so you spend them on the people who are messy and effortful and worth it. A companion that gives you back to the world is a different thing from one that quietly stands in for it. We chose the first on purpose.
Where this goes
The near future is loud, in a good way. Voice becomes the front door to software, then to your home, your car, your tools. Video will follow, and a face will make it both stranger and warmer. The use cases that matter are the ones where talking is simply the best way to do the thing: coaching, teaching, triage, rehearsal, care, the long tail of work that was always a conversation and got forced into a form. We are building toward that on purpose, with the line about companions kept in view.
The films got the feeling right and the timing wrong. It shipped. We named ours Antone. SAM took forty years to learn to breathe. Now that it has, the question is what we want these machines to be when they talk back.
If you are putting voice to work on something real, come and talk to us. Out loud, preferably.