Twenty Years Too Early

Every time I have a conversation with ChatGPT Voice, I find myself thinking back nearly three decades to a little startup in Olathe, Kansas.

It was called CompuSpeak.

There were fewer than fifty of us. Like so many startups, it didn’t feel like a company—it felt like a family. We played basketball or hacky sack in the parking lot at lunch, then headed back inside to chase ideas that seemed impossibly ambitious.

CompuSpeak was best known for voice-enabled software in the medical field. Hospitals around the country used its technology to help physicians create reports simply by speaking. Behind the scenes, we were also developing voice-driven systems for industrial applications.

We believed people would one day talk naturally to computers.

That belief shaped everything we were building.

A computer you could talk to

I was a QA Software Engineer, and one of the projects I worked on was a wearable computer for industrial inventory management. Workers wore a belt-mounted computer with a keypad and a wired headset. The unit communicated wirelessly with backend systems while voice became the primary way to interact with the computer.

I remember our developers building the software in Visual C++ and Visual Basic, and experimenting with early speech-recognition technologies—including Dragon—as they searched for better ways to make computers understand people.

Today, none of that sounds extraordinary. In the late 1990s, it felt like we were building the future.

The factory floor

One project took us into a General Electric appliance manufacturing facility. The goal was simple: keep workers’ hands free while they received instructions and responded by voice.

On paper, it made perfect sense. Reality was another story.

A factory floor is a brutal place for speech recognition. Conveyor systems, motors, compressors, forklifts, alarms, and sheet metal all compete with the human voice. We tested, refined, and tested again, but the technology simply couldn’t overcome the environment.

The vision wasn’t wrong. The timing was.

Looking back, I think we were about twenty years too early. The processors weren’t powerful enough. Wireless technology was still young. Speech-recognition algorithms struggled with noisy environments. We were asking late-1990s computers to solve problems that today’s AI handles dramatically better.

From a startup to Sprint

CompuSpeak eventually filed for Chapter 11. I was one of the employees laid off during the company’s final restructuring before it ultimately closed its doors. Like many of us, I landed somewhere else.

For me, that somewhere was Sprint.

Overnight, I went from a tiny startup of fewer than fifty people to one of the largest telecommunications companies in America, with nearly 50,000 employees. I became a Software Engineer working on early web technologies—Java, Netscape Enterprise Server, Sun Microsystems hardware, corporate intranets, and projects with budgets our little startup could only dream about.

It was exciting. It was technically impressive. But it was never quite the same.

At CompuSpeak, everyone knew everyone. Ideas moved across the room instead of through layers of management. We were building something together, even if we didn’t yet have the technology to make the dream fully real.

A promise fulfilled

Fast-forward nearly thirty years. Now I can have a natural conversation with ChatGPT Voice while driving down the highway. It understands me over road noise. It responds almost instantly. The experience is so seamless that it is easy to forget how impossibly difficult this problem once was.

When people describe today’s AI as arriving out of nowhere, I can’t help but smile.

It didn’t.

It stands on decades of work by thousands of engineers, researchers, designers, QA engineers, product teams, and dreamers who kept pushing forward long before the hardware, software, and algorithms were ready.

CompuSpeak was one small chapter in that story.

Every generation builds a piece of the bridge. The next generation carries it farther.

To the engineers, researchers, designers, product teams, and everyone at OpenAI: thank you.

Watching ChatGPT Voice feels less like witnessing a new invention and more like watching a promise finally fulfilled.

Sometimes innovation doesn’t fail.

Sometimes it is simply…
Twenty Years Too Early.