Inventing Teleports

A New Age of AI Search & The Future of the Internet’s Business Model

The fundamental laws of physics on the internet are changing. For decades, finding things on the internet was dictated by a similar user flow: typing a query into a search bar and then navigating through the results to find an answer. This core behavior is at the heart of how we interface with the internet and how search engines became the front door to the digital economy. The internet was manufactured around this behavior and the advertising model of the internet was facilitated by the search engine as a perpetual magnet for eyeballs and user intent. 

The search engine has been the primary means for transportation across the internet – getting you from point A (query) to point B (answer) by directing you from query to “10 blue links” and then from webpage to webpage to find what you’re looking for. A useful analogy is to think of the user as a car traveling from point A to point B. The benefit of this traditional transportation method is that it offers opportunities for companies to put “billboards” on each of the roads all along your journey. The foundational principle of the advertising industry is that everyone wants a well placed billboard on a frequently travelled road.  The “car” model of transportation across the internet is what has enabled the “billboard” industry to become the de-facto business model for the internet. 

The release of ChatGPT in November of 2022 catalyzed the start of a paradigm shift in how we traverse across the internet. It was the first teleport. It takes the user from point A to point B instantaneously, drastically altering the way users find information online. Teleports, or “answer engines” as they are frequently referred to, have redefined the traditional search experience by providing immediate, concise responses without the need for users to sift through multiple links and pages. More teleports have been built in the past few years – Perplexity, MultiOn, OpenAI’s SearchGPT are all well capitalized companies seeking to capture a piece of this seismic market and become a new gateway to the internet. Under increasing pressure from this group of potential usurpers, Google themselves have recently incorporated their own teleport, “AI Overviews,” into their search engine — a box that directly answers the user query using generative AI.

The user demand is clear and the bar has been set — the future of search will be defined by answer engines and ‘agentic’ search. 

The Internet’s Business Model is Breaking

As unimaginable as it has been for the past 20 years, Google’s core business is at an inflection point. Since the company’s inception, there have been two opposing forces at the heart of Google’s business. The core teams that drive revenue for Google are (1) Search and (2) Ads. The search team and the ads team are, and always have been, fundamentally misaligned. The search team optimizes for efficient search - a faster car - but the ads team optimizes for ad viewership – more billboards. In other words, the ads team is incentivized to build a user experience that encourages more queries while the search team optimizes for a faster, more efficient search, aka less queries. This paradox now faces a reckoning due to the teleports of the last few years, with Google being forced to choose between these two philosophies, potentially leading to parts of Google’s business being cannibalized to a scale never seen before. This is the direct opportunity that the teleports being built today are chasing after. Perplexity, OpenAI, Anthropic, MultiOn, Arc, Exa are all chasing this opportunity to go after arguably the greatest business model known to man - search. As these companies scale, the next decade will redefine the way in which humans interface with the internet.

As the first wave of teleports have started seeing widespread consumer adoption, expectations have been firmly set for the future of search to match, and exceed, the efficiency and time-to-value that teleports provide. At this point – this all seems great. Billions of consumers can access information faster and more efficiently than ever before. There is, however, a glaring problem – the ad model of the internet breaks in this future. Going back to our car analogy, now that we’ve invented teleports, no-one is on the roads anymore, and the billboards are becoming obsolete. The traditional advertising model, which thrived on the visibility provided by search engine results pages and website traffic, will be uprooted. This atrophy is starting to be felt across the ecosystem already today — clickthrough rates are decreasing, ROAS is decreasing and CACs are skyrocketing. 

The digital economy relies heavily on search engines as magnets for user intent and eyeballs. Many businesses are dependent on ad revenue generated from search traffic and on converting users from sponsored content on search engines. Simply put, people aren’t clicking through to websites anymore, painting a potential future where websites themselves become obsolete. This presents a profound shift on two fronts. Firstly, Google makes 60% of its revenue on search ($170B) – the proliferation of answer engines like SearchGPT and Perplexity that drive user queries away from Google directly threatens this. Secondly, countless businesses that have built their business models around search-driven traffic and ad revenue face an existential crisis. Without the traditional flow of users clicking through search results, their online visibility, revenue streams and profitability are significantly threatened. 

On the other side, there persists an open question as to how many of these teleports will monetize, with the injection of ads into LLMs being the commonly cited answer. Given the scale of the advertising industry, this seems like more of a bandaid on a broken leg than an enduring solution. We are instead presented with a unique opportunity to rewire how value is transferred across enterprises and users participating in the digital economy. The first step in this rewiring is establishing effective and equitable connectivity between AI companies and the resource they need to be effective teleports — the internet’s data.

The War for the Internet’s Data

Heavily dependent on advertising revenue for survival, content publishers have been the first to seek protection for their business model. LLMs need training data to ensure quality of model outputs and rely heavily on the internet's wealth of content for this. We are in the midst of a war of attrition between teleport creators and data providers to reach a resolution for data licensing that is deemed equitable for all stakeholders. It was ugly at first – large lawsuits were leveled at OpenAI from The New York Times for copyright infringement and a separate suit from eight newspapers claiming that OpenAI was using their proprietary data, content that they’d spend millions creating, to train their models. Last month, Condé Nast sent a cease-and-desist to Perplexity that demands the AI startup stop using content from its publications—which include The New Yorker, Vogue, and Wired—in its AI-generated responses and search results. It’s become clearer that LLM scrapers are an abrasive approach to model training and neither a sustainable nor equitable solution to improving model performance. The simple truth is that there has to be a way to monetize high quality data that enterprises have spent significant time and money curating. 

We’ve now seen the first wave of data licensing agreements that seek to forge a precedent for the future of the internet’s value chain. Reddit’s S1, ahead of its IPO, was a unique glimpse into how the new paradigm is going to affect enterprise balance sheets. Reddit signed data licensing agreements with Google and OpenAI, valued at around $66M per year for three years. To put this in relative terms – 10% of Reddit’s revenue is going to come from this new monetization channel. This is, of course, a non-trivial and significant new way for Reddit to derive value from a powerful suite of assets that have traditionally been misunderstood on public company balance sheets – their data. 

Reddit’s data is inherently valuable because it represents a vast, diverse, and highly engaged user base that generates unique, real-time content across a wide range of topics. It makes sense that the large players in the foundation model space targeted Reddit as an initial foray into structuring effective and equitable data licensing agreements – even more so when you learn that Mr. Altman owns 8.7% of Reddit. However, there are millions of websites and troves of data on the internet – deals like the one brokered between Reddit and OpenAI do not scale effectively given the majority of websites do not have the infrastructure for this data transaction to occur efficiently. An important question arises: What if every company in the world could make a significant amount of their revenue from data licensing? 

Data sales upend the current business model of the internet and stand in contradiction to how website interfaces exist today. In this instance, the philosophy behind Reddit’s product decisions will shift towards creating product experiences that produce more valuable data to feed to LLMs instead of being driven to optimize ad revenue, as it is today. The look and feel of many websites will drastically change in the next decade as a result of this mechanism occurring at scale. The team at Coframe are setting a strong precedent for the future of generative user interfaces, giving a compelling glimpse into what websites will feel like at the end of the decade. 

The LLM vendors should pay more for better data, especially as the war between foundation model providers continues to put pressure on finding new, efficient and scalable data acquisition strategies. User data remains of paramount importance, but it’s packaged & sold in a very different way. As the long awaited death of 3rd party cookies reaches its climax, their departure necessitates new mechanisms for equitable transport of user data across the internet. We have made strides over the past 5 years in building technology that provides rails for digital identity and digital ownership – through cryptography.  With the right regulatory framework, blockchain technologies will play an important part in solving a subset of the problems the internet faces today, most notably protecting trust on the internet, with companies such as Story Protocol leading the way. 

In this future, the revenue model of the Internet will have changed — from the ad model to a data sales model that can exist in tangent to the new, teleport model of transporting across the internet. As this transition starts to be felt across company balance sheets, a second important question arises: What if the revenue from data sales eventually surpasses the revenue from ads? This would present new opportunities for start-ups to build infrastructure for this to occur. Taking this one step further, perhaps instead of search engine optimization (SEO), there will be new mechanisms that are built to facilitate data sales optimization – a potential new gateway for delivering ads. It is a profound trail of thought that leads us back to Larry and Sergei’s garage for one last question: is there an opportunity to rewrite the PageRank algorithm for the agentic internet? 

The Decade Ahead

All of these forces point to a fundamental truth – the value chain of the internet will be rewired in the next decade. The only enduring solution to rebalancing the value chain involves ensuring that users keep efficiency gains from answer engines and the digital economy is still able to thrive. As agentic search continues to improve and our new gateways to the internet become more powerful, the breakdown of the ad model of the internet will force a fundamental transformation in how businesses generate revenue and engage with users online. 

This presents a profound set of opportunities for new entrepreneurs to build infrastructure for the Age of Teleports and redefine the search experience for billions of users. At Kyber Knight, we’re excited to have backed two companies in the space already: MultiOn, a browser-based AI agent that enables consumers to perform tasks on the internet completely end-to-end using natural language, and Common Tools, an early stage start-up working on the new era of personal computing. 

We’re actively investing in entrepreneurs seeking to redefine the search experience. If this is you, I’d love to chat: you can find me at ary ‘at’ kyberknight.com