A Conversation with General Magic’s Andy Hertzfeld

This month, we conclude the interview we started last issue with General Magic software wizard Andy Hertzfeld. As we noted last month, Andy has the admirable trait of speaking quickly, while at the same time saying many very interesting things.

At the end of part I, Andy had just described how the idea of General Magic had evolved inside Apple Computer. As we pick it up, Andy is about to describe how the company came on its own.

— Part II —

The initial reaction to Magic Cap for many people is to note its reasonable similarity to Apple’s HyperCard, designed by Bill Atkinson. Hertzfeld addressed this by saying: “We wrote the original prototype in HyperCard because you practically could start anything in HyperCard, especially a project that has Bill Atkinson on it.”

Hertzfeld explained: “We used HyperCard to come up with some of the basic user interface ideas of the Telecard—by March of 1990 we had something that looks very much like the Telecard we’re shipping today. The rubber stamp idea we had by March of 1990.”

However being a high-level tool, HyperCard also imposes a set of limits on developers. “So July, 1990 we left HyperCard. The only way to design an excellent user interface is through prototyping. No designer is smart enough to do in all in their head and it’s clearly an iterative process. At any given time—at least the way I like to work, the way I always worked, and the way Bill taught me how to work—you always have something running.”

Hertzfeld applies this principle religiously. He noted: “For me, even on the first day I’m working on a five year project, I’ll have something running. And it always represents the best way you know how to do it at any particular point in time. You also have to be willing to throw away 80 to 90 percent of the code you write.”

I work really hard and everyday I do that next amount of work knowing that it might have to be thrown out. But it gets me to the point where I can see the right thing to do. I just do that for 1500 days in a row and create something like Magic Cap.

Andy continued: “You write knowing that it won’t survive to the final product, but you need to write it in order to learn what the next step is. You have to step at a certain height to see where to go next. I work really hard and everyday I do that next amount of work knowing that it might have to be thrown out. But it gets me to the point where I can see the right thing to do. I just do that for 1500 days in a row and create something like Magic Cap.”

Of course, it also helps to have an energetic team of, as Hertzfeld describes it, “brilliant young programmers” around you to realize your vision.

Spinning Out

Following the April 4, 1990 board meeting where Apple initially turned down the idea of a separate business unit, Porat and company found a way to spin out. Hertzfeld recalled that: “We went through a couple months of negotiation with Apple and we founded the company on May 1, 1990. By the end of June 1990, we had our spin off agreement with Apple and it was just the three of us in a small office in Palo Alto.”

Much of the initial funding for the new company came from the three principals themselves. “We initially had to put up two million dollars between the three of us to found the company which, luckily, we were able to do. Myself from Radius and Thunderscan, Bill because of HyperCard, and Marc because he had started a prior company called Private Satellite Network.”

Hertzfeld continued: “So we had the financial wherewithal, but our asses were on the line when we did that. The money I put up was a significant fraction of all the money I had in the world. Immediately also, we knew that the key to our vision was developing an industry standard that had to be supported by many large companies.”

It’s no small ambition to want to create an industry standard. For a small company like General Magic, this not only meant teaming with the giants of the industry, but also competing to get mind share away from other projects such as Apple’s Newton.

Hertzfeld explained: “At the time we were approached by Sony and Motorola, we were completely different than the Newton. The Newton was not a communicator, they didn’t have the word ‘communication’ yet, and it wasn’t until a point after our relationship was established that they decided to change what the Newton was.”

The Thing’s a Joke…

Refusing to pull punches, Hertzfeld assessed the situation by saying: “In fact, when we started, the Newton was an $8000 computer intended for office workers, with that lousy Hobbit processor. I immediately could analyze that and see the thing was a joke.”

When asked why, Hertzfeld launched into an explanation. “Its basic architecture is not for people who want to optimize their code, the individual programmer doesn’t have enough control with the stack-based architecture. But mainly because of the economies of scale, it was a four or five chip set to get the functionality—its cost was way out of line, its power was way out of line, as were the tools.”

Summing up, Hertzfeld concluded: “There was nothing good about it, except AT&T was paying people tens of millions of dollars to use it. That was the only reason to use it.”

Envisioning Intelligent Communication

The rumor has it that early on, General Magic wanted to do communication, but wasn’t quite sure how to go about it. Hertzfeld clarified that story by saying: “We knew some aspects, we had the basic concept of the Telecard. But we didn’t have the concept of the entire intelligent network until Jim White came.”

Hertzfeld continued: “I won’t say we didn’t know how to do communications. We had this discussion with Sculley while we were still part of Apple about how Apple was going to make money off of this thing. Finally we realized the razor and the blades thing about doing a service, and we had that basic idea of doing a service while we were still at Apple.

“Once we had that idea, in one weekend I wrote a little service that enabled us to send Telecards back and forth from our HyperCard mock-up. We had that basic notion of making a communicator and actually doing the communications from the very beginning. What we didn’t have was the notion of customizing the network, or being able to put your wishes and desires into the network.”

In other words, the idea of the agent and the network serving as a platform. Hertzfeld noted: “That came completely from Jim White. And that was very exciting.”

Excitement Abounds

As we noted in last month’s leader describing the Sony Magic Link, Magic Cap’s bright and friendly icons light up the device in much the same way as in the early Macintosh computers. Hertzfeld was quick to concur.

“I’d agree with that. I think that one of the reasons you have the feeling of the Macintosh about this is the spirit behind it. It was the same spirit as the Apple II that we grew to love and cherish at Apple, and that we put into the Macintosh, and that we put into this.”

Hertzfeld continued: “It’s because it’s our concept of how to do a great product. A great product not only has to be affordable, useful, easy enough to use to be accessible, but it’s got to have that intangible spirit which is the sense of infinite possibilities, a sense of joy, that’s really the key word. You’ve got to get the sense that designers loved designing it, which makes the users love using it.

“You kind of put the exuberance in it. I think that’s really the main thing that you see. The thing is very pleased at being itself, which was true about the Apple II and the Macintosh as well. I also think that the Newton user interface is very poor, given the approach they took, mainly because they didn’t have enough time. They were always just rushing and rushing because they had to restart so many times.”

Understanding the Design Center

Hertzfeld stressed the importance of shifting the design center of the device away from a strict marketing study, to a more abstract and accessible theme. Hertzfeld explained that: “The key goals of Magic Cap include one I’ve already talked about, which is the ease of use, and to me the most exciting thing—the notion that you can reach so many more people with your work if you make it more accessible.”

He continued: “So we knew from the very beginning we had to make something an order of magnitude more accessible than the Windows, menus, mouse type way of using the computer. In fact, something about just using a touch screen creates a sense of intimacy that the mouse doesn’t have. With the mouse you’re one level removed, while here you have that feeling of direct manipulation.”

Hertzfeld described some of the design difficulties in making a device so accessible. “The finger can occupy an area of nearly 64 square pixels, so you’re very constrained as a designer—it makes it harder.”

Designing a new model for interaction is hard work and, by its nature, involves a lot of backing-up at times. Hertzfeld described the unrelenting importance that General Magic placed on feedback from user testing, a system that provides a mechanism of natural selection that filters out the weaker ideas. Being good often means discarding early interface ideas and efforts.

Hertzfeld recalled: “We discarded lots of them. OK, I’ll tell you a specific one. You know how it says up here ‘Go back to the desk.’ We thought of that as step back and for a while we didn’t have a name for where it was going. But we had a little staircase icon and we even had one of the four stairs highlighted to show which level you were at in the navigation hierarchy.”

Hertzfeld continued: “We felt that is was pretty good, except that people didn’t understand it in our user testing. When we put the name there, it made it so much clearer. And so that was a step along the way.”

Hertzfeld swears by user testing and the humility that it inevitably brings to any designer. “It’s the only way to make a good user interface. As I say, at any time have something working that reflects your best idea at the time. But don’t be proud. Don’t think the ideas are good because they’re not, and just try them on as many people as possible.”

You kind of put the exuberance in it. I think that’s really the main thing that you see. The thing [Magic Cap-based device] is very pleased at being itself, which was true about the Apple II and the Macintosh as well.

Hertzfeld noted that: “Before shipping Magic Cap, we ran somewhere between 300 and 500 people through. Every single time we had to make some change in the software. It’s very, very obvious when the user interface isn’t working; it’s very, very subtle as to why. We couldn’t have designed Magic Cap without that constant feedback of user testing.”

Hertzfeld also stressed the importance of being able to personalize any device that becomes such an integral part of your life. Hertzfeld observed that: “As you’re using it, you see your life getting poured into it, whether you like it or not.

“And it creates this more personal relationship with it—this thing is you in some sense. It’s very, very important to make it yours, to be able to change just about everything about it. And so we pursued that in Magic Cap to an extent that I don’t think the industry has seen before.”

Creating a Market

General Magic is like many other companies in that they envision the big pay-off from mobile computing coming somewhere further down the road. The major difference is, however, that General Magic has done everything to structure itself for that time frame.

Personally, Hertzfeld is optimistic about the value of these mobile devices. He conjectured: “I can envision the time when the volume of these [PDAs] is considerably higher than for desktop computers, because it will be much more likely to be used by everyone than a desktop computer. I certainly think that they will very soon be supplanting the desktop computers as where the most exciting action is in the industry—the area of highest growth.”

When it comes to hard numbers, however, Hertzfeld is somewhat cautious. “I think within three years you’ll start seeing very high growth, maybe three to five years. It’ll be more than that, perhaps seven or eight years before volumes are comparable with the desktop computers.”

To Hertzfeld, the factors and precursors are clear. “One catalyst will be getting the price down to where it’s really affordable and getting the form factor where it can always be with you—where it fits inside your pocket. Solving some of the other limitations, such as battery life will be important as well. But the key thing is breaking that $300 or so price point and getting so that it can easily fit inside your pocket.”

“I also believe it’s going to be a little bit unexpected where the slope of the adoption curve is going to get very steep at some point. The reason for that is the personal communicator that you buy is more useful to you every time someone else buys one. There’s this exponential nature, this device becomes more useful to you if I have one. And so, each one that’s sold gives everyone else a reason for buying one even more. That’s true because I use it to communicate with you, but it’s also true because of the services. The services are an enormous chicken and egg problem.”

But what about today. Is nine hundred dollars too much to pay for a Sony Magic Link communicator? Hertzfeld replied: “Yes and no. In the days of the Apple II, the question was ‘when is it a good time to buy a computer, they’re getting so much cheaper—should I wait?’ And the answer always was ‘now is the best time!’

“Because even though they could be half the price next year, you still should buy one now because it’s worth the money. The money you pay now will improve your life enough and be useful enough to you that it’s worth it today. So I would say that $900 is a bargain in some sense for getting something like this, but sure, if you wait a year it’ll go down.”

Which raises a related question: who’s the target for the machine right now, and is it the same person you designed it for when you started doing the prototypes?

Hertzfeld replied: “Well, not really. You know that there’s the early adopter phenomenon which we were familiar with through the Macintosh and other consumer electronics. So we know the early customer will tend to be a technophile, a technology optimist. But that’s not who we designed it for because if you design it for those people it just stops.”

Hertzfeld continued: “You get to them and then that’s all you can reach. So it doesn’t make sense to design for your early customers. In fact it’s a platform phenomenon. We were involved in creating the Macintosh platform, and we’ve seen other platforms be born and die, and the basic nature of a platform means you have to design it for the future.”

“That’s the reason that new platforms are always underpowered when they first come out. You saw that in the 128K Macintosh and Windows, which, when it first started could barely run on the machines it was running on, and took a full five years to really get a hold in the market. They are always like that because the platform lives over a period of ten years to make it worth anything—it’ll take a few years just to build up to anything at all.”

“You have to design the platform for the way you think things are going to be in five years. That’s just the nature of the platform game. And in five years, Moore’s Law tells you that microprocessors will be something like four or five times faster. So you have to put your design center so that it lives in the future space. That’s what we did with Magic Cap and I do believe that the way we constructed the software on Magic Cap will be the way mainstream software is constructed around the end of the century.”

Handwriting Not Ready for Prime Time

Part of building a strong foundation for a platform involves selecting technologies that work from the ones that don’t. General Magic applied this thinking to the decision not to build the platform around handwriting recognition.

Hertzfeld explained: “It was really Bill driving it. It was a matter of intuition as much as anything else. The state of handwriting recognition that was demonstrated to us—the best the world could do in the time frame of the summer of 1990 when we made the call—was that we did not think that in five years it would reach a state of being acceptable to the user.”

He continued: “I think one of the underlying reasons for that is people use their entire wealth of common sense knowledge and experience that’s been building up for a whole life to recognize every single word as you read. Context is so important—it’s a very difficult problem. But it was mainly that every single one that we investigated was a complete joke. I mean anyone who wasn’t involved in creating the technology, if they looked at it soberly, would say this isn’t even close to being usable.”

“I was on the GO advisory board in 1988 before I started with General Magic, and I was appalled when they showed me their handwriting—it didn’t even come close to working. They said: ‘Don’t worry, we’ll make it better.'”

But hold on a second. Doesn’t this conflict with what Hertzfeld said about building a platform for the future instead of getting everything exactly right the first time? Hertzfeld countered: “That’s true, but you have to have your foundations incredibly solid. You can’t build a platform on shaky foundations. You have to make a call about every single technical element. For example, the object-orientation at such a low level that we put into Magic Cap, is that going to be too slow to build a practical system? You might think yes, you might think no, you have to make a call.”

Hertzfeld continued: “Handwriting recognition is very much the same thing. Or speech recognition which we considered but thought no, it would kill us if we tried to base it around speech recognition even though that will come. So the basic way it works has to be solid early. You can’t design the rest of it if your basic way just doesn’t work, if you can’t count on it.”

More Opinions…

Having asked about the Newton, it’s only fair to solicit Hertzfeld’s opinion about the Zoomer. Hertzfeld’s response was swift: “I think the Zoomer makes Newton look great (laughs). I can’t say anything good about it, not a single thing.”

Hertzfeld shrugged-off the news that the AST version was recently selling for $160 at the Costco warehouse club by saying: “Well, I think that’s because no one bought them. I know the design center is fairly low cost, but to me, let other people do the work of taking existing stuff and making it cheaper. That doesn’t require so much imagination, maybe at certain levels it does.”

Hertzfeld expanded on this by saying: “I think that thing was designed for the past, in terms of a new philosophy. It was worst than the status quo the day it came out. So, I don’t see any value in that, there was no innovation. Whereas the Newton at least had some spirit and some touches that I liked.”

Toys R Good

Hertzfeld was particularly quick and impassioned when responding to the common criticism that Magic Cap is too toy-like. “Yes, exactly right. When the Mac came out, one of the first reviews described the control panel as a toy, and we took that as a complement.”

I think toys and tools are the two most fun things to make, because there is nothing more fun than bringing joy into someone’s life… So if you can make something that’s both a tool and a toy, I can’t think of anything finer to do.

“So I think toys are great—I think toys and tools are the two most fun things to make, because there is nothing more fun than bringing joy into someone’s life. And a tool has extraordinary leverage, it’s something that empowers people. So if you can make something that’s both a tool and a toy, I can’t think of anything finer to do.”

Hertzfeld concluded: “I take something being called a toy as essentially a complement. Now, if people said the user interface doesn’t work, it’s gratuitously cute—I saw one review use the word saccharine—I don’t like that. And part of it is a lack of vision or, I don’t know if the right word is Calvinist or something like that, some people believe that anything that’s good for you has to hurt.

“That’s the opposite of what I think. So I think that if you give people a few years, they’ll come to see that it’s good to design things to be fun.”