Marc Andreessen

The Network Is the Computer: Google Tries Again


All you need is a dumb device attached to a smart network. It’s an old idea that refuses to die despite repeated failures. Now it’s Google’s turn.

In the late 1980s, Sun Microsystems used a simple, potent war cry to promote its servers: The Network Is The Computer. Entrust all of your business intelligence, computing power, and storage to Sun’s networked SPARC systems and you can replace your expensive workstation with a dumb, low cost machine. PCs are doomed.

Nothing of the sort happened, of course. Sun’s venture was disrupted by inexpensive servers assembled from the PC organ bank and running Open Source software.

PCs prospered, but that didn’t dampen the spirits of those who would rid us of them.

Fast-forward to the mid-1990s and thought re-emerges in a new guise: The Browser Will Be The Operating System (a statement that’s widely misattributed to Marc Andreessen, who holds a more nuanced view on the matter). The browser will serve as a way to access networked services that will process your data. The actual OS on your device, what sort of apps it can run — or even if it can run any (other than a browser) — these questions will fade into insignificance.

Soon after, Oracle took a swing at the Network is the Computer piñata by defining the Network Computer Reference Profile (or NCRP), a specification that focused on network connectivity and deemphasized local storage and processing. It was understood, if not explicitly stated, that an NCRP device must be diskless. A number of manufacturers offered NCRP implementations, including Sun (which would ultimately be acquired by Oracle) with its JavaStation. But despite Larry Ellison’s strongly expressed belief that Network Computers would rid the industry of the evil Microsoft, the effort went nowhere.

Today, The Network Is The Computer lives on under the name Cloud Computing, the purest example of which is a Google Chromebook running on Chrome OS. (And thus, in a sense, Sun’s idea lives on: Google’s first investor was Sun co-founder Andy Bechtolsheim.)

So far, Chromebooks have shown only modest penetration (a topic for musings in a future Monday Note), but despite the slow adoption, Google has become one of the largest and most important Cloud Computing companies on the planet. Combine this with the Android operating system that powers more than a billion active devices, could Google bring us to the point where The Network Really Is The Computer?

It’s a complicated question, partly because the comparison with the previous generation of devices, traditional PCs, can (excuse me) cloud the view.

Unlike PCs, smartphones rely on an expensive wireless infrastructure. One can blame the oligopolistic nature of the wireless carrier industry (in English: too few companies to have a really competitive market), but that doesn’t change the simple fact that wireless bandwidth isn’t cheap. The dumber the device, the more it has to rely on the Cloud to process and store data, and the more bandwidth it will consume.

Let’s visit Marc Andreessen actual words regarding Network-As-Computer, from a 2012 Wired interview [emphasis mine]:

“[I]f you grant me the very big assumption that at some point we will have ubiquitous, high-speed wireless connectivity, then in time everything will end up back in the web model.”

If we interject, on Andreessen’s behalf, that wireless connectivity must be as inexpensive as it is ubiquitous, then we begin to see the problem. The “data hunger” of media intensive apps, from photo processing to games, shows no sign of slowing down. And when you consider the wireless bandwidth scarcity that comes from the rapid expansion of smartphone use, it seems that conditions are, yet again, conspiring against the “dumb device” model.

The situation is further confounded when we consider that Google’s business depends on delivering users to advertisers. Cloud computing will help drive down the cost of Android handsets and thus offer an even wider audience to advertisers…but these advertisers want a pleasant and memorable UI, they want the best canvas for their ads. When you dumb down the phone, you dumb down the ad playback experience.

In a recent blog post titled The next phase of smartphones, Benedict Evans neatly delineates the two leading “cloud views” by contrasting Apple and Google [emphasis mine]:

“Apple’s approach is about a dumb cloud enabling rich apps while Google’s is about devices as dumb glass that are endpoints of cloud services…”

But Google’s “dumb glass” can’t be too dumb.  For its mobile advertising business, Google needs to “see” everything we do on our smartphones, just like it does on our PCs. Evans intimates as much:

“…it seems that Google is trying to make ‘app versus web’ an irrelevant discussion – all content will act like part of the web, searchable and linkable by Google.”

Native apps running on a “really smart” device are inimical to Google’s business model. To keep the advertisers happy, Google would have to “instrument” native apps, insert deep links that will feed its data collection activities.

This is where the Apple vs. Google contrast is particularly significant: iOS apps are not allowed to let advertisers know what we are doing – unless explicitly authorized. Apple’s business model doesn’t rely on peddling our profile to advertisers.

In the end, I wonder if Google really believes in the “dumb glass” approach to smartphones. Perhaps, at least for now, The Computer will remain The Computer.


The Browser Is The OS: 19 Years Later


So it was declared in the early days: Web apps will win over native apps. Why let the facts cloud an appealing theory?

Marc Andreessen, the Netscape co-founder, is credited with many bold, visionary claims such as “Everyone Will Have the Web” (ca. 1992), “Web Businesses Will Live in the Cloud” (1999), “Everything Will Be Social” (2004, four years before joining Facebook’s Board), and “Software Will Eat the World” (2009).

But not all of Andreessen’s predictions are as ringing and relevant. His 1995 proclamation that “The Browser Will Be the Operating System” still reverberates around the Web, despite the elusiveness of the concept.

The idea is that we can rid our computing devices of their bulky, buggy operating systems by running apps in the Cloud and presenting the results in a Web browser. The heavy lifting is performed by muscular servers while our lightweight devices do nothing more than host simple input/output operations. As a result, our devices will become more agile and reliable, they’ll be less expensive to buy and maintain, we’ll never again have to update their software.

The fly in the ointment is the word connected. As Marc Andreessen himself noted in a 2012 Wired interview [emphasis mine]:

[I]f you grant me the very big assumption that at some point we will have ubiquitous, high-speed wireless connectivity, then in time everything will end up back in the web model.

So what do we do until we have ubiquitous, high-speed wireless connectivity?

We must build off-line capabilities into our devices, local programs that provide the ability to format and edit text documents, spreadsheets, and presentations in the absence of a connection to the big App Engines in the Cloud. Easy enough, all you have to do is provide a storage mechanism (a.k.a. a file system), local copies of your Cloud apps, a runtime environment that can host the apps, a local Web server that your Browser can talk to… The inventory of software modules that are needed to run the “Browser OS” in the absence of a connection looks a lot like a conventional operating system… but without a real OS’s expressive power and efficiency.

For expressive power, think of media intensive applications. Photoshop is a good example: It could never work with a browser as the front end, it requires too much bandwidth, the fidelity of the image is too closely tied to the specifics of the display.

With regard to efficiency, consider the constant low-level optimizations required to conserve battery power and provide agile user interaction, none of which can be achieved in a browser plug-in.

Certainly, there are laudable arguments in support of The Browser Is The OS theory. For example: Unified cross-platform development. True, developing an app that will run on a standardized platform decreases development costs, but, let’s think again, do we really want to go for the lowest common denominator? A single standard sounds comfy and economical but it throttles creativity, it discourages the development of apps that take advantage of a device’s specialized hardware.

Similarly, a world without having to update your device because the Cloud always has the latest software is a comforting thought.. but, again, what about when you’re off-line? Also, a growing number of today’s computing devices automatically update themselves.

In any case, the discussion may be moot: The people who pay our salaries — customers — blithely ignore our debates. A recent Flurry Analytics report shows that “Six years into the Mobile Revolution” apps continue to dominate the mobile Web. We spend 86% of our time using apps on our mobile devices and only 14% in our browsers:

Apps 86 Browser 14

…and app use is on the rise, according to the Flurry Analytics forecast for 2014:

Apps Web Flurry 2013 2014

So how did Andreessen get it so wrong, why was his prediction so wide of the mark? It ends up he wasn’t wrong… because he never said “The Browser Will Be the Operating System”. Although it has been chiseled into the tech history tablets, the quote is apocryphal. 

While doing a little bit of research for this Monday Note, I found a 1995 HotWired article, by Chip Bayers, strangely titled “Why Bill Gates Wants to Be the Next Marc Andreessen”. (Given Microsoft’s subsequent misses and Marc Andreessen’s ascendency, perhaps we ought to look for other Chip Bayer prophecies…) The HotWired piece gives us a clear “asked and answered” Andreessen quote [emphasis mine]:

“Does the Web browser become something like an operating system?

No, it becomes a new type of platform. It doesn’t try to do the things an operating system does. Instead of trying to deal with keyboards, mouses, memory, CPUs, and disk drives, it deals with databases and files that people want to secure – transactions and things like that. We’re going to make it possible for people to plug in anything they want.”

Nearly two decades later, we still see stories that sonorously expound “The Browser Is The OS” theory. Just google the phrase and you’ll be rewarded with 275M results such as “10 reasons the browser is becoming the universal OS” or “The Browser Is The New Operating System”. We also see stories that present Google’s Chrome and Chromebooks as the ultimate verification that the prediction has come true.

The Browser Is The OS is a tech meme, an idea that scratches an itch. The nonquote was repeated, gained momentum, and, ultimately, became “Truth”. We’ll be polite and say that the theory is “asymptotically correct”… while we spend more energy figuring out new ways to curate today’s app stores.

On Marc Andreessen’s optimistic view of news


A strongly-worded column by venture capitalist Marc Andreessen triggered an intense debate on the future of news. Andreessen might be right places, but his views can also be dangerously simplistic. 

For starters, it is always great to have an outsider’s view. Marc Andreessen’s witty, and fast-paced dithyramb on the future of news is undoubtedly welcome. But, as always, regardless of the depth and breath of the big picture he paints, the devil lies in the details. In no particular order, here are my thoughts on his manifesto.

As a European, I found his piece extraordinary US-centric or, slightly more broadly, Anglophone-centric.

Andreessen wrote :

[T]he market size is dramatically expanding—many more people consume news now vs. 10 or 20 years ago. Many more still will consume news in the next 10 to 20 years. Volume is being driven up, and that is a big, big deal.
Right now everyone is obsessed with slumping prices, but ultimately, the most important dynamic is No. 3 – increasing volume. Here’s why: Market size equals destiny. The big opportunity for the news industry in the next five to 10 years is to increase its market size 100x AND drop prices 10X. Become larger and much more important in the process.

By saying this, Andreessen makes two good faith mistakes.

First, he mixes up global reach and monetizable audience. Evidently, a growing number of people will enjoy access to news (maybe not all the 5 billion cellphone users he mentions), but the proportion of those able to generate a measurable ARPU is likely to be very small.

The Scalability that works for Google Maps or WhatsApp doesn’t work as well for the notion of relevant information, one that is more tightly connected to language, proximity and culture.

Second, he overestimates the addressable news market’s fragmentation. I live in France, a 66 million people country with a high standard of living and good fixed and mobile internet access. In spite of these factors, it remains a small market for the super-low-yield digital news business that brings few euros per year and per user (except for a minuscule subscriber base.) I remained stunned by the inability of good journalistic products, created by smart people, to find a sustainable business models after years of trying.

And the huge, globalized English speaking market does not warrant financial success. The Guardian is one such example. It operates one of the finest digital news system in the world but keeps bleeding money. The Guardian brings a mere $60m in digital ad revenue per year — to be compared to a kitten-rigged, listicles-saturated aggregator generating a multiple of this amount. Journalism has become almost impossible to monetize by itself (I’ll come back to that topic).

Andreessen also vastly underestimates the cost of good journalism when he writes:

[T]he total global expense budget of all investigative journalism is tiny —  in the neighborhood of tens of millions of dollars annually.”

Fact is, journalism is inherently expensive because it is by laborious and unpredictable: An investigation can take months, and yield nothing; or the journalistic outcome can be great, lifting the reputation of the media, but with zero impact on the revenue side (no identifiable growth in subscriptions or advertising). The same goes for ambitious coverage of people or events. No one has ever translated a Pulitzer Prize in hard dollars.

This is also the case for what Andreessen calls the “Baghdad Bureau problem”. It was said to cost $3m/year for the New York Times. In fact, on an annual basis, the Times spends about $200m for its news operations, including $70m for foreign coverage alone. The NYT is likely to stay afloat when it goes entirely digital (which might happen before the end of the decade), but one of the nastiest features of digital news is the unforgiving Winner Takes All mechanism.

As far as philanthropy is considered, I won’t spend too much time on the issue except to say this: Relying on philanthropy to cure malaria or to support ill-understood artists bears witness to an absence of sustainable economic system. (Until, perhaps, the artist dies; as for malaria, there is indeed a very long term benefit for society, but not for those who supply the treatment, hence the mandatory call to generosity.) Saying investigative or public-interest journalism could/should rely on philanthropy is the same as admitting it’s economically unsustainable. Luckily, American society has produced scores of philanthropists free from any agenda (political, ideological, religious) — such as the Sandler Foundation with ProPublica. That’s not the case in France — not to mention Russia and many other countries.

There are plenty of areas in which I completely support Marc Andreessen’s view. For example: A media company “should be run like a business“, i.e. seek the profitability that will warrant its independence (from every economic agent: shareholders, advertisers, political pressure, etc.) This brings us to the size and shape of a modern news factory (I use the term on purpose). We have to deal with an unpleasant reality: Good journalism is no longer sustainable as a standalone activity. But — and that’s the good news — it remains the best and indispensable core around which to develop multiple activities (see my recent column about The News Media Revenue Matrix).You can’t develop services, conferences, publishing, etc. around a depreciated journalistic asset. On the other hand, this asset has to be drastically streamlined: In many cases, less people, better-paid (simply for the ability to retain talent) and with sufficient means to do their job (don’t go for the press junkets because the travel budget has been slashed, you’ll lose on three counts: credibility of your brand, self-esteem of your team, quality of the reporting.)

Unfortunately, as Andreessen noted, there are plenty of hurdles to overcome. In fact, most existing news companies do not fathom the depth of the transformation required to survive and thrive. Nor do they understand the urgency to set this massive overhaul in motion. Such moves require strength, strong leadership, creativity, a fresh approach, unabated confidence, and a systemic vision — all of the above in short supply at legacy media. Note that when Marc Andreessen prides himself to be an investor in media ventures (for instance Business Insider– no conflict of interest), all are digital natives and bear none of the burdens of traditional media. His bullishness on news is selective, personal.