How Unlimited Money Killed the First Smartphone: The General Magic Story

In 1994, a Silicon Valley company shipped a pocket device that did most of what an iPhone does. Email. Calendar. Contacts. Animated messages you could send to a friend on the other side of the country. A built-in network of services. It was beautiful. It was thirteen years before the iPhone.

It also failed. Less than two years after the launch, the network behind it shut down, and its partners soon stopped making devices.

The company was called General Magic. The people who built it went on to build the iPod and the first iPhones, Android, eBay, Nest and the software of the Apple Watch. They were not stupid. They were not lazy. They were one of the most talented product teams Silicon Valley ever assembled.

So why did they fail?

The honest answer is uncomfortable for anyone who believes that better people, better technology and more money produce better outcomes. General Magic had all three, more than anyone in its market. And the abundance is precisely what killed it.

The note

Around 1989, Marc Porat, then at Apple, described a pocket device in a note to John Sculley. The project was code-named Pocket Crystal. The text is famous in Silicon Valley, partly for what it predicted and partly for its language. “It must be beautiful,” Porat wrote. He compared the device to a piece of jewelry, a seashell and a crystal, and insisted that it had to be something other than just another telephone.

In May 1990, Apple spun the project out as a separate company, General Magic, and kept a stake in it. Porat led the company. Bill Atkinson and Andy Hertzfeld, two of the engineers behind the original Macintosh, came along as co-founders. Susan Kare, who had drawn the original Mac icons, designed the interface. Joanna Hoffman, the original marketing strategist of the Macintosh, ran marketing.

Over the next few years, the team grew to include a young engineer named Tony Fadell, who would later lead the iPod and help build the first iPhones; Andy Rubin, who would later co-found Android; Megan Smith, who would later become Chief Technology Officer of the United States; Pierre Omidyar, who would later found eBay; and Kevin Lynch, who would later lead the software of the Apple Watch.

The money

Sony and Motorola came in first. Then AT&T, Matsushita and Philips. By 1994, the alliance around General Magic counted sixteen companies, from NTT and Toshiba to France Telecom.

They were not just investors. They were partners who committed to building hardware and offering network services around General Magic’s software. Sony would build a device. Motorola would build another one. AT&T would run the network.

The partnerships meant General Magic did not have to worry about distribution. It did not have to worry about manufacturing. It did not have to worry about telecom infrastructure. It did not even have to worry about market education, because some of the biggest brands in the world were lined up to do that work.

Its sixteen backers put in about 200 million dollars. In February 1995, General Magic went public. The stock nearly doubled on its first day of trading, while the first device had been on sale for barely five months.

On paper, General Magic had won. The remaining question was how big the win would be.

Building for a customer who did not exist yet

For four years before the launch, the team worked around the clock. Some engineers slept under their desks; others built bunk beds in the office. They built two ambitious pieces of software from scratch: Magic Cap, the operating system of the device, and Telescript, a language that let small programs, called agents, travel across the network and act on behalf of their users.

What they never pinned down was the customer. The official target was “Joe Sixpack”, a caricature of the average American on his couch. Nobody took it seriously: engineers designed the features they liked and assumed Joe would like them too. In the documentary, a team member sums it up: they thought everyone would want one because they did, and they were wrong. Much later, in his book Build, Tony Fadell wrote that even if Joe Sixpack had existed, he was never going to buy the Magic Link.

This was not an oversight. It was a philosophy. They were building a category, not a product. The customer was an abstract person in the future who would, on first contact with the device, understand instantly that this was what life was supposed to be.

If you have read any lean startup book published after 2010, you already know what is coming. But in 1990, this philosophy was respectable. Steve Jobs was famously hostile to market research. The idea that you might call a few hundred people and ask them what their actual problem was felt amateur, almost insulting to the work.

So they kept building. For four years.

The launch

Sony launched the Magic Link on September 28, 1994, at a list price of $995. Motorola’s Envoy followed in early 1995, at about $1,500. Both ran Magic Cap. Both connected to AT&T’s PersonaLink network, which cost extra. Both did, technically, much of what Porat’s note had promised five years earlier.

Sales never took off. The device was expensive, and most of the things it did were things people did not yet want to do on a pocket device. Communication was supposed to be the killer application, but few people had a device, and PersonaLink never reached 10,000 subscribers.

Less than a year after the Magic Link shipped, Netscape went public. The web was about to swallow the world. A proprietary network suddenly looked like an expensive curiosity next to a free, open alternative.

AT&T shut down PersonaLink in August 1996. By 1997, the partners had stopped making Magic Cap devices. Magic Cap got a last life as a Windows 95 application. The company turned to voice services over the phone, then shut down in 2002.

According to the documentary General Magic, fewer than 3,000 devices were sold.

The diaspora

If the story ended there, it would be a footnote. A clever device, ahead of its time, killed by the web. The interesting part is what happened to the people.

Tony Fadell went to Philips, started his own company, then joined Apple in 2001, where he led the team behind the iPod and the first iPhones. He then co-founded Nest, which Google bought for 3.2 billion dollars.

Andy Rubin co-founded Danger, the maker of the Sidekick, then Android, which Google bought in 2005 and which now runs on more than three billion devices.

Pierre Omidyar launched AuctionWeb, the future eBay, in September 1995, while still working at General Magic.

Megan Smith ran PlanetOut, spent eleven years at Google, including at Google[x], and served as Chief Technology Officer of the United States under President Obama.

Kevin Lynch led development at Macromedia, became chief technology officer of Adobe, then joined Apple, where he led the software of the Apple Watch.

Susan Kare went on to design for Facebook, PayPal and Pinterest.

Many members of the team built something on the scale of an industry. The thing they could not build was General Magic itself.

The curse of capital

Here is where the story becomes useful.

Common explanations for General Magic’s failure point to bad luck, bad timing, or the arrival of the web. These explanations are comforting but incomplete. The team had four years, about 200 million dollars and some of the smartest people in the industry to see the web coming. They missed it because they were not looking outside.

The harder, more useful explanation is this: General Magic failed because it had everything.

When you have a strong founding team, you do not have to convince anyone to join, so you never have to explain clearly to an outsider what you are building.

When you have an Apple pedigree, you do not have to prove anything in the market to raise money, so you never collide with the discipline of selling something.

When you have the biggest names in electronics and telecoms as partners, you do not have to find your own customers, so you never learn who they are.

When you have 200 million dollars, you do not have to make a smaller version of your product, so you ship the whole vision at once and have nothing to fall back on when it misses.

When you have four years before you have to launch, you do not have to face the market during the build, so you arrive at launch day fully committed to assumptions that have never been tested.

Each of these advantages, on its own, looks like a good thing. Together they form a system. The system removes every forcing function that would normally push a startup to talk to customers, launch small, kill its weakest features, and accept that its vision is partly wrong.

This is the curse of capital. Money does not just buy you time. It buys you time without the pressure to use it well.

The same pattern inside large companies

Thirty years later, I still see this pattern in the companies I work with. Initiatives get bigger and slower as more money is poured in.

This happens because the project becomes too critical to the company’s future. Executives want to ensure it succeeds, so they give it more resources.

It never works for projects without customer traction. Teams are protected from the market, often forbidden from talking to customers. Nobody enters the room to ask hard questions. The budget keeps flowing and the program keeps growing, until it fails to deliver revenue.

More funding does not fix that. It gives the team more time to build what nobody asked for.

What lean startup was built to prevent

The lean startup method, formalized by Eric Ries in 2011 but rooted in the work of Steve Blank in the 2000s, is a direct answer to the General Magic problem.

Build the smallest thing that tests your riskiest assumption. Ship it to a small group of real users. Measure what they actually do, not what they say. Learn. Iterate. Repeat.

If you find those sentences obvious, it is because decades of expensive failures have hammered them into the industry. They were not obvious in 1990. They sounded like the kind of advice you give to someone too poor or too unimaginative to do the real work.

What lean startup does, mechanically, is reintroduce constraints on purpose. A minimum viable product is a constraint. A two-week sprint is a constraint. A customer interview is a constraint. A pivot meeting is a constraint. Each one forces a small, painful decision that a well-funded team can postpone forever.

This is why the method works best in startups with limited resources, and why it tends to die quietly inside large companies that try to “do lean” with unlimited budgets and no consequences for being wrong.

Constraints are not a limitation. They are the mechanism that creates customer discipline. You cannot fake them. If you have unlimited time, money and political cover, you will use them, no matter how good your intentions are. The General Magic team had the best intentions in Silicon Valley. They still spent four years building a device almost nobody bought.

The practical lesson

If you run a startup or a new product team today, the General Magic story is not just a cautionary tale. It is a checklist of warning signs.

You are at risk when your founding team is so strong that nobody asks whether the market wants what you are building.

You are at risk when you raise so much money that you can afford to delay launching.

You are at risk when your partners are so big that your reputation does not depend on customer adoption.

You are at risk when your vision is so complete that any single piece of feedback feels like an attack on the whole.

You are at risk when, after a year of work, you cannot name three real people who have used your product and told you something painful.

If you find yourself in any of these positions, the right move is not to keep building. The right move is to manufacture constraints. Set a shorter deadline than the budget allows. Ship something smaller than the vision demands. Ask ten people, in person, whether your problem is really their problem. Cancel a feature this week. Make life harder for yourself on purpose, because the alternative is to drift gently, for four years, toward a launch day where nobody shows up.

The question to ask this week is not “what else can we add?” It is “what is the minimum we have to ship for one customer to actually use this?” The answer is not in your roadmap. It is in the head of a real customer somewhere.

General Magic had every advantage a startup can have. The advantages were the problem. The discipline they lacked was not a method. It was the willingness to confront customers, in real life, before the product was ready to be praised.

That willingness is what lean startup is, underneath the diagrams. And it is what separates the teams who build the iPhone from the teams who, with the same talent and more money than they need, build the Magic Link.

Sources: the documentary General Magic by Sarah Kerruish and Matt Maude (2018); New York Magazine, an oral history of General Magic (2018); Computer History Museum, Computing in Your Pocket: The Prehistory of the iPhone in Silicon Valley (2017); General Magic on Wikipedia; Tony Fadell, Build (2022), quoted by Commoncog, and his interview with The Verge (2022).