Everyday Chaos

by David Weinberger

  There’s more to the story, though. The Upshot is hosted by GitHub, a site used by developers around the globe to enable teams to collaborate fluidly and agilely. Because developers can post code to the site in a way that is publicly available, people can reuse or alter it without first asking permission. Furthermore, if you reuse a part of someone else’s open-source code in your own project, GitHub remembers that and makes the relationship public. If the original developers incorporate a change that you’ve made, GitHub records that reuse as well. This not only gives credit where it’s due, it creates an ecosystem for the sharing of code. GitHub thus has become an enormous upstream repository of code and code fragments that can be put to use in entirely unanticipated ways: in September 2017, GitHub had twenty-eight million users and hosted over eighty-five million projects.37

  Arfon Smith, chief scientist at Github when I spoke with him, told me that although GitHub was originally built for software developers, “the platform turns out to be for collaboration in general.”38 According to Smith, by 2014 people were using GitHub to share and improve on knitting patterns, to gather home repair procedures, to organize weddings, and to write jointly authored papers. “People are forking cocktail recipes,” Smith said with delight, referring to the process of altering someone else’s code in order to create something new.

  GitHub began by allowing developers not to anticipate how their code could be used. By refusing to anticipate even what type of projects might benefit from unanticipated upstream sharing, GitHub’s utility has reached far beyond the world of software development.

  * * *

  As the success of all these techniques shows, the true price we paid for the old way of doing things became apparent once we had tools and technology that let things happen differently.

  Preparing for Spontaneity

  Embracing unanticipation doesn’t mean that we must bid adieu to all prediction and planning.

  “The event was designed with a funny combination of persnickety attention to some details, and a sanguine letting go of others,” says Sara Winge, cocreator and producer of the first of Tim O’Reilly’s Foo Camp “unconferences.” (This is the same Tim O’Reilly behind the “government as platform” idea.) “I do think we hit on a pretty good system that takes care of enough stuff—both physical infrastructure and social norms—that the participants can get a sense of what’s going on and also understand that they have a responsibility to make wonderful and/or meaningful things happen.”39

  Getting there meant the O’Reilly organization had to have many discussions about whom to invite, how many simultaneous sessions should be allowed, whether to put whiteboards in the discussion areas, whether on the first night there should be an all-attendee ice-breaking exercise, how big the grid of available rooms and times should be, and whether attendees should write directly on the scheduling whiteboard or use sticky notes instead.

  Winge remembers one particular discussion vividly. She argued strongly that the name badges should not list attendees’ affiliations, even though the group was eclectic and thus contained many strangers. She acknowledged that not listing affiliations would make some people “cranky,” but “it reinforces the point that the event is about the people who are there,” not the organizations they work for. Plus, it was a bit of social engineering: “If you want to know where somebody works, you have to ask them.”

  In short, Foo was an event carefully planned to enable unplanned discussions.

  We have been learning that, more often than not, enabling the unanticipated requires this type of thoughtful anticipation. Simply opening up an API or publishing open-source software and hoping people will use it often doesn’t work. For example, many of the news media’s APIs have been underutilized, and some have been closed. So long as humans do X in order to gain Y, unanticipation is never going to replace the old anticipate-and-prepare strategy entirely. We’re still going to resupply our larders. The local theater company is going to print up more than enough programs for the attendees. Businesses are still going to stock inventory and plan for the holiday rush. But there is nevertheless something new in the world: we are increasingly willing to give up control in order to enable the emergence of things of value that we didn’t predict.

  We are thereby learning that creating a future that’s even further out of our control can be a surprisingly productive strategy.

  The Platform for Platforms for Unanticipation

  In their 2010 book The Power of Pull, John Hagel, John Seely Brown, and Lang Davison argue that open platforms are an essential part of the major shift business is undergoing from “push” to “pull.” “Push operates on a key assumption—that it is possible to forecast or anticipate demand.”40 Pull, on the other hand, attracts contributors from outside the business and gives them the tools—often in the form of open platforms—that let them extend and repurpose products. The authors of The Power of Pull consider this to be an epochal change that transforms not just economies but also the production of knowledge and culture overall. Likewise, the technology writer Kevin Kelly considers open platforms to be one of the “12 inevitable forces that will shape our future,” as he says in The Inevitable’s subtitle.41

  Beneath all of these changes is one of the most profoundly unanticipatory technologies in human history.

  That the internet was architected for unanticipation is apparent in one of its foundational documents: “End-to-End Arguments in System Design,” published in 1984 by Jerome Saltzer, David Reed, and David Clark.42 This technical paper makes the counterintuitive argument that the best network provides the fewest services. For example, it might have seemed obvious to the internet’s early designers that its users would need a search engine. But if the network designers built one into the system, it would address only the needs that the designers could anticipate. It’d be far better to enable any user to build a search engine. That way competition can work its magic, producing far better search engines than any centralized committee of network engineers could manage. It means niche engines can arise. The same is true for security services, archiving, and the other services one can anticipate users of networks would want.

  As Reed put it to me, “Optimizing for one service de-optimizes for others.”43 That’s the problem with overanticipating what your product is for and building it to address that use. It is also why Reed has been a strong advocate for a policy of net neutrality that forbids internet access providers from deciding what they think the internet is “really” for—delivering the movies their parent companies own, perhaps?—and optimizing it for delivering those bits ahead of others.

  Because the net was held back from anticipating its uses, it became useful for everything, from reading research papers, to making phone calls, to watching videos, to holding collaborative drawing sessions, to taking online courses, to controlling our household appliances, to whatever is the next thing someone invents.

  We now have a generation that has grown up with the internet as a fact of life. They take it for granted that if they have something they want to share, they can post it. They take it for granted that if they have an idea for a service to offer over the internet, they can just build it. These are core assumptions for them. But they are assumptions that imply the power of a system that refuses to anticipate and prepare for what it might be used for.

  Every time we touch the net, we relearn the same lesson: unanticipation creates possibilities. It means we no longer need to pay the heavy price of wasted resources or missed opportunities that come from over-, under-, or mispreparing. More importantly, instead of limiting the value of what we build by anticipating and preparing for the few narrowed-down possibilities that we could foresee, we are now building to meet needs that a connected world of users might invent for one another. By anticipating and preparing, we were not only gambling that our idea of a customer’s first automobile was right but also making it harder—often virtually impossible—for users to add value to our products by extending them, adapting them, an
d customizing them for their own workflows.

  We had little choice about the matter. Anticipating and preparing was the right strategy until digitization made our products more extensible, and our new global network connected users who were eager to collaborate and share what they’ve made.

  The effect goes beyond increasing customer satisfaction and loyalty. Being able to make more of a product than we anticipated lets the world show itself to us in a new light. “This is for that” becomes a needlessly limited way of thinking. Yes, this was intended for that, but it could be for this, for that, or for something no one has thought of yet. And if we’re learning that often there are serious benefits to holding off on locking things into their anticipated uses, then how things interact is also freer, more possible, and more complex than we’d thought.

  Indeed, as we will see in the next chapter, in a world that bursts the bonds of anticipation, everything affects everything. All at once. Forever.

  Coda: Libraries of Anticipation

  Open Text, a developer of searching software … and Yahoo Corp., publisher of a comprehensive directory to the on-line universe, are now sharing their technologies.44

  This was news in 1995, and very big news for Open Text, where I was vice president of marketing. We were a small company with a great text search engine, and Yahoo! was the queen of the internet; from the early days of the web, it was where you went to find stuff. As PC Magazine put it at the time, “Every day, Yahoo serves up over 6 million pages to 400,000 users.”45 That’s what dominance looked like back then.

  Before the incorporation of the Open Text search engine, if you wanted to find information about, say, art therapy, at Yahoo! you’d scan the fourteen categories on the home page and choose Health. Clicking on it took you to a new page with a couple dozen subcategories. Choose Mental Health, then Therapies, and then Arts Therapy, and at last you’d be looking at links leading to eight pages on the topic.46 Each of those links had been chosen by hand by Yahoo!—originally by the site’s two creators, grad students at Stanford—and individually placed into its tree of topics.

  If you knew what you were interested in, browsing through a tree of topics was far better than just poking around the thousands of pages on the web at that point, but it was not nearly as efficient as searching. That’s why it was a big deal when Yahoo! decided to put a search box right at the top of its home page. Now you could put in the words you were looking for and the site would find pages that talked about, say, art therapy, no matter what branch they were hanging from.

  The matchup of browsing versus searching was decisively settled a few years later when two other Stanford grad students created a search engine they whimsically named Google—as whimsical as “Yahoo!”—that kicked browsing’s butt. Yahoo! pivoted to become a media and entertainment site, struggling as Google bestrode the Earth like a colossus.

  Anticipation is at work in both searching and browsing, just as it is for any purposive action we humans undertake, but searching’s anticipations are very different from browsing’s. We search when we have a narrow enough idea of what we need for our project: “Prius 2007 headlight change” or “Dallas vegetarian barbecue restaurant with jukebox.” The very nature of a search box encourages us to be highly specific about what we want.

  We browse when we’re asking a question that has many right answers. “What would I like to read?” is very different from “What is the capital of Peru?” The right response to a search is one-dimensional: the end point of a straight line, the house at the end of the street, the light at the end of the tunnel—a right answer. Browsing requires as many dimensions as we can manage—dimensions not as physical spaces but as pathways through a set of possibilities. In a library, the physical shelves arranged by topic create one dimension, but users also have the ability to browse by author, by strong reviews from other readers, by new acquisitions, by books on sale for one dollar, by librarians’ recommendations that are purposefully just different enough from our usual fare that we are surprised and then delighted.

  There is an art to creating a place to browse, a game of overpreparing and unanticipating. The library has anticipated that we might enter to look for books, and the clothing store has anticipated that we’re there to find something to wear. But while both searching and browsing involve anticipations and intentions, browsing is defined by its relative lack of anticipation. Browsing is to searching as the weekend is to the week.

  That means libraries are in the business of being overprepared because they cannot finely predict all the works that their communities might want. Only about 3 or 4 percent of Harvard University Library’s magnificent collection is checked out every year, and it’s not a completely different set of works each year, so there are more than a few books that have not been checked out within living memory.47 Imagine a restaurant that keeps an item on the menu that hasn’t been ordered in over fifty years. It takes almost a thousand librarians to run Harvard’s library system, with an annual budget of over $150 million, but it’s worth it to the university because it enables the faculty and students to find answers to questions that they did not know they were going to be asking, and that may literally never have been asked before.48

  In fact, a browsing space like a library needs noise: choices we can consider and reject, choices we can skip over. For any individual user on any particular trip to a library, well over 99 percent of it will be noise. Done right, books in a library are going to be rejected so frequently that if they were young actors who just arrived in Los Angeles, they’d be on the next bus back to their hometowns, brokenhearted.

  Being overprepared is expensive, but when it comes to the ingredients of creation, not of consumption, it is a necessary gift. Granted a free rub on a genie’s lantern, most librarians would prefer to create libraries that are so overprepared that they could provide every resource to every user. Someday perhaps the law will permit what technology already enables: making all resources freely available online under reasonable copyright constraints. (The United States started with copyright lasting for fourteen years; we’re now at seventy years after the death of the author, and climbing.) Until then, the extravagance of possibility libraries offer announces their commitment to resolutely not anticipating precisely what their community should be interested in.

  Meanwhile we see all over the internet the growth of collections of information where the economics of anticipation have flipped: it is now usually cheaper to include digital content than to exclude it. That’s why site after site lets users upload whatever they want without permission. If there are legal issues or abuse, that can be weeded out afterward. And if there’s too much stuff uploaded higgledy-piggledy—which is to say, if the site succeeds—the service may need to come up with better ways to let users filter content on the way out, rather than resorting to filtering it on the way in.49 The result is collections the value of which is vastly amplified by the fact that they were not limited by the curators’ anticipation of what users will use them for. What is true of the open collections on the net is true of the net itself.

  The internet has taught us many lessons, and one is that anticipation doesn’t scale. Another is that anticipation constrains possibilities. Unanticipation liberates them, the way a great library does.

  Chapter Four

  Beyond Causality

  Interoperability

  We don’t expect things to happen on the internet the way they happen in the real world. Not only is the pace of change exponentially faster on the internet, it also gives us a direct experience of unpredictability every day. We check to see what the latest quirks are online and off. We see tiny internet sparks ignite worldwide flames. We see ridiculousness paraded for our entertainment or scorn. We prowl to see what new thought has been thought and what new invention has been invented. Just as a vessel flying through outer space never knows what piece of space dust, drawn to some large but distant mass, might breach its hull, a web browser flying through internet space never knows if it
’s going to come across a video game—Pokémon—controlled by the simultaneous but conflicting flicks of hundreds of users, or a page with dots of dust whose path across the screen you can alter by clicking to make “gravity points” wherever you choose.1 There seem to be no rules guiding change on the internet, and we have come to accept that, to count on it, and to like it.

  In chapters 1 and 2 of this book, we also saw how machine learning is giving us a model of models so complex that we often cannot fully comprehend them. What happens, we are learning, is the result of everything that is happening, all at once. While our new technology lets us predict more accurately and makes predictable domains we formerly thought were too complex, it simultaneously reveals the immense complexity we skate over.

  We are so far down the path of this new model of change that we take for granted minimum viable products, open application programming interfaces (APIs), and so much more, even though they fly in the face of tens of thousands of years of anticipating the future and preparing for it.

  It turns out that all these changes in our understanding of how the future happens are rooted in a single but profound change in our most basic model: it no longer resolves to the laws of causality. Beneath causality, we are discovering something even more fundamental.

  Plug-in World

  If you need to buy an audio cable to plug your mobile phone into your car’s stereo system, you can get one at your local dollar store and it will work in any car that has an auxiliary plug. But if you need a replacement bulb for the car’s headlights, you may end up paying over a hundred dollars, and not just because headlights are more expensive to manufacture than audio plugs. The car manufacturer has you over a barrel because you need a bulb designed for that one type of car. You are, in effect, paying the noninteroperability tax.