The era of the shapeshifting web is upon us

published on Apr 06, 2024

Until now, the most common way to interact with generative AI has been through text-based prompts. Drawing a parallel, just as computer user interfaces transitioned from text-based Command Line Interfaces to graphical ones, I believe we might witness a similar evolution with generative AI.

In fact, generative AI might play a crucial role in ushering in the next iteration of the web: the true Web 3.0.

(“But,” you might interject, “we already have a Web 3.0.” To that, I’ll provide my response at the end of this article.)

To understand why this might be the case, let’s briefly revisit past transitions. Bear with me, as this won’t take long, and you’re likely already familiar with this history. However, it’s important to review it to logically build upon it.

From Web 1.0 to Web 2.0

Web 1.0: The Television for Documents

When the web first appeared, the real innovation was the ability to link to documents from within other documents. This was essentially the extent of the interactivity that came with what later became known as Web 1.0.

The client’s only agency was to choose which page of predefined content to navigate to next. In essence, the web at that time was akin to a television for content: you could change the channel (i.e., page), but you couldn’t alter how the shows looked.

Web 2.0: It’s… Alive!

The advent of Web 2.0 marked a significant shift: dynamic HTML. In other words, this meant that web pages could be built on the fly and display bespoke content for each user.

In practice, this allowed developers to build websites where two users could visit the same page (for example, and see completely different content. To achieve this, servers had to pull pre-recorded data from a database and insert it into predefined HTML templates.

This same technology also meant that a more significant change could now take place: platforms could be built for and around user-generated content.

Over time, the tools for Web 2.0 became more sophisticated and complex, but the core principle remained the same: project data from a database into a dynamically generated template. This principle continues to underpin the structure and functionality of many of today’s most popular websites.

Setting the ground work

One aspect that has been continuously optimized is the process of determining what data to display to a specific user. These processes have grown more complex over time, taking into account an increasing number of factors. At some point, artificial intelligence was introduced.

Step 1: AI as a mediator

At the beginning, AI was tasked to be a mediator between the database and the template. The server continues to pull data from a database and display it within a template. However, the method of data retrieval has shifted from being purely deterministic to incorporate more probabilistic aspects.

This approach still aligns closely with the Web 2.0 model.

The key difference is that past user behavior now influences the algorithm's decisions on what data to pull. To those not keeping track of how these things work, the involvement of artificial intelligence might not be apparent.

Step 2: AI as a creator

The advent of generative AI has introduced a different type of user experience. In this case, the displayed data isn't waiting in a database for the server to retrieve and display. Instead, the server takes a direct and specific instruction from the user to generate content that is, to some extent, unique.

This still resembles Web 2.0, albeit less so. We continue to use templates to display content, but instead of pulling the data from a database, it is generated by an algorithm.

The leap forward

All of this is to say that there are some advancements on the web axis, although they are relatively small. Up until this point, we only altered two main elements that describe the web in its 2.0 incarnation: how data is generated (by AI instead of by users) and how that data is pulled (again by AI, instead of deterministic algorithms).

Some might label this as Web 2.1, while others might see it as closer to Web 2.9. It's almost the next step, but not quite there yet.

What's missing?

The only part where the AI isn't involved yet, at least in a user-facing manner, is in how the data is presented. Until now, this was done using pre-made templates. Yes, more of these templates are now developed using AI, but that's not the point. It's the predefined part that is key here.

If AI models mature enough, it won't be a stretch to see websites whose very architecture is adapted to the user's needs in (near) real-time. The content can be static and stored in databases, but the way it is presented can drastically change from one user to the other.


Thus, the articles on will still be written and meticulously fact-checked by the Times reporters. But instead of reading their coverage on the 2024 solar eclipse, you'll ask the website to generate a podcast — and by a podcast, I don't mean someone just reading the article.

Imagine, instead, the article transformed into an episode of The Daily where Michael Barbaro discusses the piece with the reporter who wrote it. Or maybe as an episode of Office Ladies. Or maybe you want it as a video in the style of your favorite YouTube creator, Mustard.

In all these instances, the content will stay exactly the same, but the article written and edited by professional journalists will serve as the basis for AI-generated content in other formats.

(As an aside, you’d also need to pay the Office Ladies or Mustard a fee to use their voices/styles/etc., either as a pay-as-you-go or as a subscription. This represents a potential new business model for content creators in the era of AI and Web 3.0.)

...and shapeshifting

The website may also shapeshift to better serve you its content in the format you desire. So it may look like a podcast app if you prefer audio, or... even as a Super Mario game if that's what you desire.

All generated by an AI. So all the developers at The New York Times would have to provide are the API to fetch the (base) content, and a way for the "readers" to interact with the AI model, either via raw text-based instructions or a more simplified button-based interface.

Or maybe you want to master the TCP/IP stack. You go to the excellent TCP/IP guide but there's just too much to learn. No problem! The new website, built using web3 tech, will help you learn in a breadth-first manner, allowing you to soak in the new knowledge layer by layer (pun not intended!), avoiding the usual abandonment mid-course (again, not intended).

But if you think that's still pretty boring, you can ask the AI to transform the course into a "The Office" version where Michael Scott teaches the Dunder Mifflin employees all about TCP/IP.

Just as today, two people visiting the very same URL can see totally different content, those using Web 3.0 will experience completely unique interfaces when using the exact same website.

The possibilities are limitless.

And the cherry on the cake: the amazing backward-compatibility that's baked right into the web, means that sites built for different versions can all live together side-by-side and help keep the internet teaming with whatever the internet is teaming with!

But what about the decentralized Web 3.0?

Given the length of this article, I’ll keep this brief. It would have been more fitting, and potentially even more impactful, to name that particular tech stack “The Internet 2.0”.

This name suggests a complete overhaul or upgrade of the entire internet, not just the web. The term “Internet 2.0” would encompass advancements beyond the web, including improvements in infrastructure, protocols, speed, and perhaps even the incorporation of new technologies like quantum computing.

The term “Web 3.0” implies an evolution of the web as we know it, but the blockchain’s vision goes beyond that. It’s not just about creating a new type of website or improving web-based applications. It’s about changing the fundamental way in which data is stored and transactions are conducted online. It’s about creating a decentralized network where data is not controlled by any single entity and transactions are transparent and immutable.