How We Lost The Web — The “Big Missing Feature” according to Marc Andreessen

7 min readJan 15, 2021

One of the top VCs in the world is saying that the world would have turned out differently if users had been able to annotate everything — to layer knowledge on top of web pages.

What in the world would change?

This is the essence of what the Overweb Challenge is. Join us in changing the future of the web.

The Internet is a digital Wild Wild West. Most places (other than social media) on the web are ghost towns. Other people are on the site but you can’t see them. And then there’s social media. You don’t know who or what is real, or who is being real.

What if you could see and interact with people on websites?

Hence, quoting information on the web is like sitting on an old chair. You never know if it’s gonna hold.

You have to be careful what you believe online. Anything goes on the web with respect to expression, as long you express within the constraints of the respective platform (e.g., 280 characters). There is no filter, no validation, no peer review, no cross check, little evidence cited, and few links provided.

When We Lost the Web

It was not always that way. The web started out as communities of researchers exchanging knowledge within their field. They used links to connect and share information… and there was no advertising.

Researchers loved the notion of layering knowledge upon web pages. We call this annotation. Some of the earliest annotations are in the early Jewish texts of the Talmud, which not only cite each other, but comment upon one other. Medieval rabbis both commented on the Talmud itself and on each other’s comments. A closer to home example of annotation is the notes you see in the margins of second-hand books. Many of us have annotated and may still annotate books, and therefore have a first-hand understanding of their potential value.

Annotation was a key aspect of all of the web’s predecessors: Vanevar Bush’s Memex in 1947, Douglas Englebart’s “Mother of All Demos” in 1969, and ARPANet. Annotation was also in Tim Berners-Lee’s proposal for the web at CERN in 1989 as well as a feature in the world’s first widely distributed web browser, Mosaic.

But annotation was dropped in 1995 during the transition from Mosaic to Netscape. There was no cloud to store the annotations and Microsoft was pre-loading Internet Explorer on all PCs at the time. Despite early success, Netscape had no real chance against the behemoth Microsoft who released IE 1.0 soon after Netscape. Also based on the Mosaic code base, which it licensed from Spyglass, Inc. for peanuts, Microsoft continued to improve IE until they eventually overtook Netscape, offering at that point a consistently better product than their counterpart. Netscape was bought by AOL for billions in the late nineties and shut down a decade later.

Nevertheless, back in 1995 — when Netscape dropped annotations — was when we lost the web. In 2013, Marc Andreessen, co-author of Mosaic; co-founder of Netscape; and co-founder and general partner of Silicon Valley venture capital firm Andreessen Horowitz, explained why they were investing $13M in Rap Genius, an annotation tool for rap lyrics. Among other reasons, he mentioned:

Finally, there’s the other reason — maybe even the real reason — why I’m so fired up about this idea and this investment. Only a handful of people know that the big missing feature from the web browser — the feature that was supposed to be in from the start but didn’t make it — is the ability to annotate any page on the Internet with commentary and additional information.

This is a huge deal — the co-author of the first browser said the web browser has “a big missing feature” — annotation. We might ask ourselves what are the implications of not having such a vital feature in web browsers for 25 years. And why we still don’t have that feature in web browsers.

How the Web Turned Out

Absent of annotation, advertising on free platforms that harvest user data perhaps inevitably became the dominant monetization method on the web. Although ads may have become king even with annotation, an evolution of annotation could have created many innovative and competitive monetization models over the years.

In the past decade, search and social media algorithms along with fake accounts created an Internet of silos.

Information is in silos.

Commercial sites avoid (or downplay) outbound links because they want to keep you on their site, serving you more ads and extracting more of your data. If you leave, you are lost to them. The ramifications of this are huge. For example, a news article about a scientific report that does not link to the report. A less obvious example is Facebook providing just enough information about a posted article that most users will choose to like or even share without actually clicking the link.

Hence, the web has little context, which makes it difficult to discern what is real, what is not real, and what is missing. The resulting information asymmetries make people vulnerable to misinformation and disinformation on the web.

People are in silos.

The web flattens complex beings into a two-dimensional grid and reduces them into filter bubbles. Predatory algorithms assisted by fake virality and false claims on the web channel people’s attention with the intention to change their behavior.

The purpose is generally to nudge someone into purchasing something, voting for someone, being confused, or even performing acts of violence.

This dynamic is enabled by browsers that render the Internet flat and static, while treating people and information as commodities.

One might say that we are working for technology rather than technology working for us. Some may say we are controlled if not cognitively enslaved considering our technology addiction and the extreme value exchange asymmetry. Others may say it’s more like an abusive marriage. Although we have many choices of what to do on the web, on any given Internet platform, we are constrained to a limited set of choices.

This constraining of choices works well for the AI. Each Internet platform has its own AI algorithms tuned to the choices they give users. For example, with Twitter, we can post 280 characters, follow other users, reply, direct message other users, make lists, flag content, and maintain a profile. AI algorithms work best with small numbers of “features” or variables. The less degrees of freedom for you, the easier for the algorithms to predict and influence your behavior.

In some ways, it appears we are in a rodent maze constructed of imperceptibly dynamic digital walls created by algorithms that are tweaked by the data scientists in order to monetize our behavior.

But The Internet is Not Flat

The Internet is a hyper-dimensional system that interconnects disparate electronic systems across the planet and through space. The browser — in concept — has the potential to access the Internet in its full dimensionality. But in practice, we see only a hint of hyper-dimensionality with the ever more endangered hyperlink as well as browser extension annotation tools like Hypothesis and Diigo, which enable users to add notes to text on web pages.

If browsers were to embrace the Internet in its full dimensionality, we could have a deeply connected web that layers knowledge on top of web pages. We could have an Internet of Ideas.

In the 2013 post, Marc Andreessen continued:

Back in 1993, when Eric Bina and I were first building Mosaic, it seemed obvious to us that users would want to annotate all text on the web — our idea was that each web page would be a launchpad for insight and debate about its own contents. So we built a feature called “group annotations” right into the browser — and it worked great — all users could comment on any page and discussions quickly ensued. Unfortunately, our implementation at that time required a server to host all the annotations, and we didn’t have the time to properly build that server, which would obviously have had to scale to enormous size. And so we dropped the entire feature.
I often wonder how the Internet would have turned out differently if users had been able to annotate everything — to add new layers of knowledge to all knowledge, on and on, ad infinitum.

Let that sink in. One of the top VCs in the world is saying that the world would have turned out differently if users had been able to annotate everything — to layer knowledge on top of web pages. This is the essence of what the Overweb Challenge is about.

The Overweb is a trust layer over the webpage that enables knowledge and interactions to be layered on web pages, creating a hyper-dimensional web. On today’s web, at any given point, you have at most one option. Either you can click something or not. In the hyper-dimensional web, any piece of content can spawn dozens if not hundreds or more options. Paraphrasing Andreessen, each piece of content becomes a launchpad for insight and debate about itself… and more.

In the Overweb, the web page that currently defines our experience (and enables us to be controlled) becomes the contextual footprint for interrelated layers of information, interactions, transactions, and experiences that emanate from pieces of content on the page.

The next articles in this series will examine the Overweb Pattern and tools that unleash what’s possible on the Overweb.

Similar to the Wild Wild West being the seed of Silicon Valley, one way or another, the web will become hyper-dimensional.

The Overweb Challenge is your opportunity to be in the conversation.

+++++++++++++++++++++++++++++++++++++++

The Overweb Challenge is sponsored by the Forbes Funds and the Overweb Foundation. Partners include the EU’s flagship Next Generation Internet program, Bridgit.io, UMI, Edgeryders, and Excellerent Solutions.