From Tamagotchi to Gabbo: The Full History of Interactive Toys

If you were a 90s kid, you remember the first time a pet responded to you. Whether it was your Tamagotchi holding on for dear life or your Furby speaking absolute gibberish, it felt real. To kids today, that same feeling is amplified by ten. AI toys are capable of so much more than what we grew up with, but sometimes it's worth taking a trip down memory lane, and that's what this blog sets out to do.

The 1990s: First "Real" Pet

This era was defined by responsive, animal-like pets. First, we have to address the elephant in the room: the big dog, Tamagotchi.

Tamagotchi

In 1995, Akihiro Yokoi, an executive at a company called WiZ, pitched the idea to Bandai after hearing about a boy who couldn't take his pet turtle on vacation. Bandai's Aki Maita then turned the concept into a pet-raising simulator.

Tamagotchis had three buttons: A, B, and C. You would set the timer, and within a few minutes an egg would hatch. From there, you had to feed it, play with it, discipline it, and clean up after it. How well you cared for it determined how it looked as it grew, and with enough attention, your pet would turn into a nicer-looking creature than one that had been neglected.

When Tamagotchi launched in Japan, it became an instant craze, selling close to 10 million units within its first year. Bandai had to ramp up production to nearly 3 million units a month just to keep up with demand, and the craze soon reached America, cementing Tamagotchi as one of the most recognizable toys ever made.

Furby

Furby was created by Dave Hampton and Caleb Chung. The two of them spent about a year and a half designing and creating the toy, with the goal of making an expressive electronic companion. They brought in Richard C. Levy, a fellow toy inventor, to help sell the concept, and Levy's efforts landed them a licensing deal with Tiger Electronics in 1997. Tiger Electronics was then bought by Hasbro that same year, which gave the team the resources they needed to make the idea a reality.

Furby was designed to feel like a pet you could interact with, not just a toy. Out of the box, it spoke in gibberish, but the more you played with it, the more it started mixing in real English words. That was what made it feel alive: it seemed to learn, dance, and talk back.

Furby is often described as the physical successor to Tamagotchi. Where Tamagotchi was limited to a small screen, Furby could move, dance, and physically be in the room with you.

Giga Pets

Giga Pets were essentially Tiger Electronics' answer to Tamagotchi, released around the same time and often thought of as the off-brand alternative. Even so, it was hugely influential, and plenty of people still remember it fondly.

The 2000s: Toys Get Programmable

In the late 1990s, microcomputing was getting significantly cheaper to mass-produce at the toy level. This was a massive upgrade to the "brains" inside these toys. For the first time, kids could actually control how their toy behaved, essentially turning them into little programmers.

Lego Mindstorms

LEGO had been working on the idea of a programmable brick since the 1980s, looking for ways a computer could actually communicate with and control its toys. After nearly a decade of work, that idea became the RCX brick.

The Robotic Command Explorer (RCX) was a brick containing an 8-bit microcontroller, 32 KB of RAM, and 32 KB of ROM, and it could receive instructions over infrared. The catch was that you needed a computer, Windows or Mac, to actually write the code before sending it to the brick.

This toy was really the origin point for modern programmable robots, most of which now run on AI.

Robosapien

Released in 2004 and designed by Mark Tilden, a robotics engineer who had worked with NASA, Robosapien had a humanoid shape with a distinct early 2000s robot look. It shipped with over 67 programmable functions that could be chained together into fluid sequences of movement.

One distinctive thing about Robosapien was that Tilden built it to be openly hackable, meant to be opened up and reprogrammed by enthusiasts and kids alike. The real appeal of the toy was the hope that one day we might have an actual robotic companion, and no one quite knew what that would eventually look like.

The 2010s: Voice Assistants

With the invention of voice assistants like Siri and Alexa, the major toy companies wanted to get in on the "Talking AI" train. And Mattel, the company behind Barbie, answered the call.

Hello Barbie (2015)

Mattel partnered with a startup called ToyTalk to create a Barbie you could actually talk to. That was the dream, anyway. In reality, it worked like this: a button on Barbie's belt prompted a question, and the microphone recorded the child's response and sent it to the cloud. Speech recognition software converted the audio into text, and AI software pulled out keywords from that response, triggering Barbie to reply with one of 8,000 lines written by a team of human writers.

So not exactly true AI yet, but they were getting there, and the idea of leaning on the cloud was the right one. The hard part was that ToyTalk was essentially trying to build something like Alexa or Siri, without the massive investment and years of research those had behind them. With fewer resources, the finished product came out a little rough, and the voice would sometimes stutter or repeat itself.

CogniToys Dino

A startup called Elemental Path originally built a prototype called "GreenDino" to enter the 2014 IBM Watson Mobile App Developer Challenge. They ended up winning the grand prize, which secured them a partnership with IBM. The toy was only a bare-bones prototype at that point, but that changed quickly.

After a Kickstarter campaign in 2015, they raised over $275,000, despite only asking for $50,000, to fund production and shipping. The finished toy was a green dinosaur that needed an app just to set up Wi-Fi and some basic information, but after that it stayed screenless. It talked directly to the child through the cloud, similar to the AI toys we have now. The toy itself was a genuine success, but it eventually met its downfall.

Even though its child voice recognition was impressively accurate, it fell short on actual thinking. Reportedly, plenty of questions just couldn't be answered and got filed into an "answer later" category. What finally killed it was a couple of compounding factors: the company being too small, and the cloud computing needed to keep the toy talking getting more expensive per unit every year. A couple of years later, their servers shut down for good, and the toy became a cautionary tale for every AI toy company that came after it.

The 2020s: Toys That Can Think

The last two decades led to one giant buildup toward AI toys. The capability and the vision both got propelled forward into the 2020s, and we ended up with genuinely capable toys. As we write this blog, it wouldn't be fair not to mention ourselves in the AI toy category, so think of this next part as your advertisement tax.

The Pioneer: Curio

Curio started as a small startup with our CEO, Misha Sallee, and Sam Eaton, our Chief Toymaker. We started with a vision: what if we could make a toy that spoke the way a person actually would? With AI developing so quickly at the time, that vision was starting to feel like something we could actually build.

What really fueled the idea was the ability to use a Large Language Model (LLM) to hold open-ended conversations with a child. That means no question is too tough or too niche to ask, and the toy will almost always have some kind of response. Each toy comes with Bluetooth and Wi-Fi, with controls managed through the app. That app is actually a big step forward, since toys in the 2010s usually kept all of a child's data on the cloud with no real parent access. Curio changed that, letting parents actually see the conversations happening between their child and the toy.

Curio is really the latest step after decades of buildup toward a toy that's genuinely useful, interactive, and fun.

What's Next?

Here's an honest section about the toys mentioned in this blog. Curio was the first to build a toy around a genuine large language model, capable of truly open-ended conversation, while the toys before it leaned on narrower or more scripted forms of AI. That's an important distinction, but it doesn't mean earlier toys weren't using real technology; they just weren't capable of the same kind of free-flowing conversation. And AI toys keep getting better by the day. Even comparing Gabbo V1 to Gabbo V2, the technology has gotten a serious upgrade already.

There are a lot of forces shaping the AI toy space right now, and development seems to be heading in a few different directions. First, toys are going to get much faster reaction speeds and more fluid conversations, including better interpretation of garbled toddler speech. Second, AI toys are due for a serious overhaul of their cloud computing and security, largely because of advancing technology and, fairly enough, growing public pressure.

But to be completely honest, nobody truly knows where AI toys will end up. One thing is certain: they're only going to get more advanced from here.

Conclusion/TL;DR

As the title suggests, we've gone all the way from Tamagotchis to Gabbo, and it's wild to think about how far the technology has come since then. Still, there's something nice about looking back at where it all started. We hope this brought a little nostalgia for the parents reading, or maybe taught you something new along the way. Either way, thanks for taking this walk down memory lane with us, and as always, stay curious.