Dr. Roy Tsezana describes a transition from a fixed and mobile network to the "network of agents" - networks representing users, bargaining with each other, and creating new rules for search, advertising, protocols, and security. New opportunities alongside major challenges.
There are articles that are too long and complex for the Internet. Articles that were supposed to be chapters in a book, but the world is moving so fast that by the time I finish the book – everything will have changed three more times. So this article will be here, because it is important. It is a landmark in the evolution of a new kind of Internet. And if you are smart, talented and energetic, you might even be able to become the new millionaires of the new network thanks to it.
The signals about the evolution of the new network are coming from everywhere. Newspaper sites are reporting a decline of tens of percent in the number of their readers. This is not a slight blow to the wing, but a decline that sometimes reaches eighty percent. And why? Because Google's artificial intelligence reads the news and summarizes it on its own for everyone who asks. It is likely that a similar phenomenon is also occurring for blogs. Even WhatsApp is now launching a new feature for groups, in which artificial intelligence summarizes all the correspondence in the group in recent days.
The Internet used to be a virtual space where humans surf, browse, and read. Now humans are becoming less and less of a part of the Internet, and artificial intelligence is replacing the surfers. The Internet is becoming, as a new study recently claimed, an Agentic Web – a network of agents. Not Her Majesty’s agents, but AI agents.
The implications of this change are going to be dramatic in every sense. When websites don't receive traffic from humans, they also can't show ads to surfers, and therefore have a hard time making a living. When there are no ads on websites, Google (which is responsible for a large portion of them) loses one of its main sources of income, along with giants like Outbrain and Taboola. When surfers don't go to Amazon, wander through the virtual shelves and find offers for purchases they hadn't thought of - but that they are tempted to buy - then the company loses income, and so do the small merchants who offered their wares there. And this is just a small part of the changes we will see, and which will lead to a shake-up in the business model currently accepted on the Internet.
To understand the magnitude of the transformation and the new opportunities and threats it opens up, we need to go back to the history of the web. This is because the internet has already undergone two revolutions in the past, and we must understand them before diving into the new revolution.
One last note before we begin: I am writing this article based on A review paper recently published on arXiv, and is a collaboration between researchers from eight universities in China, the United States, and the United Kingdom. Such collaboration is not common, and adds confidence to the idea of the "network of agents" that the authors promote.
And now let's begin!
The First Network Revolution: The Age of Personal Computers
In the era of personal computers, the web was based mostly on websites. Those websites were like islands – or continents, in the case of particularly large and important sites – that people would come to visit and browse. But the number of islands was growing rapidly, and each island also had cities and streets and special areas of interest to different people. Someone, or something, was needed to bring order to the whole thing.
And so Google was born.
Google introduced a search engine that allowed users to reach the most relevant sites for them. Other search engines preceded Google, of course, but Google was able to provide the most relevant results to users.
The era of personal computers was characterized by the fact that users knew what they wanted, and search engines acted on their instructions. If I wanted to find information about woolly mammoths, I would enter the query into Google – and it would recommend that I go to Wikipedia, the Encyclopedia Britannica website, the National Geographic website, and maybe a popular blog or two that wrote about the woolly mammoth. But it all started with an explicit request from me: Find me information about woolly mammoths.
This was the situation throughout most of the first decade of the 2000s, and then the next revolution came and a new era began.
The era of the mobile network
The era of the mobile web began, not surprisingly, with the introduction of smartphones into our lives. Suddenly everyone had a camera and video camera, a microphone and most importantly – a connection to the web in the palm of their hand. Everyone took pictures, videos, recorded podcasts, and uploaded everything to the web. And the web absorbed all this information and grew and grew.
Not every area of the web grew equally. News sites remained roughly the same size, but social networks exploded with content. Users didn’t even know what to search for, so search engines were replaced by “recommendation engines.” These are the artificial intelligence engines that decide what will appear in your Facebook feed, or what the next video you’ll see on TikTok will be. There was a movement from “guided search” to “lucky discovery.”
Smartphones launched the mobile web era, and they also amplified and perpetuated it. When a user reads on a smartphone, they are not interested in in-depth research. They want small, specific, entertaining, and exciting pieces of information. In short, they want a TikTok video that is no more than a minute long. The fact that you are reading this article and have made it this far is nothing short of a miracle. And if you have indeed made it this far, include the word "elephant" in the comments so that we know you did it. And congratulations to you.
The mobile age has added new challenges to our lives. Facebook, Instagram, Twitter, and TikTok largely control the information we are exposed to, as they can influence their “recommendation engines” to choose what to show us. Foreign countries can try to sneak in advertisements and false information. All advertising services have reconfigured themselves to adapt to the recommendation engines. This is what a new era of the Internet means.
And now we are entering a new era.
The era of the agent network
"I instructed my agent to arrange a flight for me to New York next week. The agent immediately went into action: He checked against all my previous flight reservations to understand what price range would suit me, and from there he went on to contact all the relevant airlines. Not with the human clerks, of course, but with their agents who were waiting online for someone to talk to them. He received offers from all of them simultaneously, and then did some research online to understand the level of each airline, and which airport in New York would be the best to land at, and even what day would be best for me to land so that I would arrive refreshed for my meeting there. After all of this, he re-contacted the agents behind the three most suitable offers, and negotiated with them to lower the price to the requested level. He gave me the final offer he received on the phone, and recommended that I confirm it. He offered to see his entire course of action and logic, but who has the power to go through all of that? I trusted him. He knows what he's doing."
Welcome to the Agent Network. It's built on the same Internet infrastructure as before, but it focuses on a new type of user: agents. That is, the AIs that surf the web like humans, talk to other AIs, and do business with them.
In the Internet as we know it today, islands and continents were adapted to humans. True, they had a certain structure that attracted search engines or recommendation engines, but as the saying goes, "content wins." That is, if you created good content, people were willing to read, listen to, or watch it all the way through, and recommend it to their friends. In the agent network, a large part of the Internet will begin to be adapted to the agents themselves.
What is meant by?
If today there are news sites for humans, then the agent network will have news sites… for agents. They will be specifically tailored to agents and their way of reading, and will try to attract agents to read them. Why? So that the agents will be influenced by the bias of the news, and pass it on to the human user for whom they work.
If Amazon is currently specifically tailored to humans, then in the agent network it will turn to agents. It will entice them to review certain products that the user has sent them to purchase, and will also try to interest them in related products that the user may not yet know they want. And of course it will adjust its prices to the omniscient agents, who are able to simultaneously review dozens of products in dozens of different stores online, and find the lowest prices. Oh, and Amazon will also have its own agents, who will bargain with the user's agents.
If today hackers on the Internet try to trick users into entering their details into fictitious websites, in the near future they will start focusing on agents. They will create websites with special instructions for agents, which humans do not see at all, but the agents read and obey. They will try to get the agents to get out of control and spend the user's money on illusions and lies. And they will create automated Nigerian princes, who will try to trick agents into transferring information or money to their account.
This will be the network of agents. A network in which most of the traffic will consist of agents running back and forth to fulfill the desires of the users who sent them. Or agents operating every minute, or every hour, or every day, conducting research, collecting information, analyzing data and writing reports or investing in the stock market. Either way, one thing is clear: most of the information flow on the network will no longer be the result of human activity. It will be the responsibility of agents.
And what about the “mobile web” – that is, the social networks where people share information all the time? It will also be full of agents. Some of them will be “one-to-many” agents. That is, they will take the place of human ‘influencers’ (or work for them), and publish messages and videos that will reach many. They will even be able to respond to their followers.
The second type of agents on the mobile network will be “many-to-many.” These will be the agents that ordinary people will activate to help them with social interactions. Want to advocate for Israel? Your agent will respond on your behalf in relevant threads. Think Bibi is king? Your agent will spread your ideology, and never stop arguing. Just keep in mind that he will mainly talk to other agents, because the other side will also give up on convincing you – and will hire his own agents to do the work.
The agent network, did we mention it?
New questions and opportunities
In the world of the agent network, new questions arise, and those who know how to answer them will discover great opportunities awaiting them.
One question is how agents can tell which sites are ‘good’ and which are not. Google’s algorithm decides which sites will appear at the top of the list based on a number of different factors, including the number of people who visit and use the site. But what happens when agents are the ones visiting the sites? Or when bad agents – those with a flawed mindset – visit the wrong sites, messing up the entire algorithm?
Understand how critical this is: Google rose to prominence largely due to its ability to distinguish between sites and pages with important and reliable information and those that are irrelevant. In the agent network, it needs to rethink the way it does this. Whoever helps it – or finds an alternative mechanism – will rise to prominence quickly.
You might say that if most of the browsing is done by agents, then there is no need to know which sites are more or less important? But if that were the case, how would the agents themselves know which sites they should visit to find the information they need? So a new product is needed: search engines and recommendation engines… for agents.
Another question, and perhaps the most important, is where the money is. Many websites are currently keeping their heads above water (barely) thanks to advertisements and cultivating relationships with their users. What happens when most of the human users disappear, and are replaced by agents? How will websites continue to exist, without a source of income?
One answer is that the sites will run a new kind of ads: for the agents themselves. But how exactly do you do that? Will agents be tempted by ads of half-naked women leaning on Mercedes? Or of gnarled men riding horses in the middle of the desert with cigarette butts between their lips? Or do you have to talk sense into the agents and convince them that they should recommend their human operators to buy them? We don’t know. It’s a blue ocean—that is, uncharted territory—waiting to be filled with ideas.
Think that's all? Not even close.
What about security? How can webmasters prevent agents from entering their sites and consuming valuable bandwidth? Agents can already handle mechanisms like Captcha with relative ease ("Are you human? Identify the bike in the picture!"). What will be the next defense mechanisms against agents, and who will want to use them?
And what about the security of the agents themselves? There will certainly be sites that will act as "traps" for agents, and try to give them instructions (in text invisible to humans). How will agents protect themselves? And if there are agents who write their own articles (and there will be many of them), how do you prevent other agents from being convinced by distorted viewpoints or recommendations that are distributed on the social network?
How can the three networks – stationary (websites), mobile (social networks), and agent – coexist? Will news sites also have to upload new articles in a format that is especially suitable for agents? Or will influencers arise that will specifically target artificial intelligence agents? Again, we don’t know. Again, there is another opportunity here for service providers and product developers in the field.
What will be the protocols for conversations between agents? Today, it is common for systems to communicate with each other via API, but this is an outdated method. In the past year, we have switched to the MCP protocol (Model Context Protocol). This protocol allows agents to communicate with each other and also explain to each other the context of their conversations and requests. Alongside the MCP, Google has just proposed a new protocol called A2A (i.e., Agent to Agent), which should help agents talk more freely. Which protocol will win? We don't know.
How will agents transfer funds? How can the user trust them to be careful with their passwords and bank accounts? We don't know.
What about energy? Advanced agents like Genspark can run dozens of sub-agents, which will operate throughout the network to fulfill the user's desires. What does this mean for human energy consumption? How do we ensure that the Internet doesn't get clogged with too much browsing volume? We don't know.
Looking a little further, what will happen when the network of agents connects with the "network of robots": robots connected to the Internet? How do you prevent private agents from operating robots that are not theirs? Or alternatively, how do you allow some agents to operate robots in certain situations, and who is responsible if the robot causes damage in the physical world?
Want to start the next big startup? Try to find answers to these questions.
Summary
In the last twenty years, we have experienced two internet revolutions: the era of the "fixed network" undermined and disrupted the way established communication works, and created some of the largest companies in the world, such as Google. It was followed by the era of the "mobile network", with Facebook, Twitter, and TikTok riding its waves. Each of these companies did not succeed on its own: it acquired dozens or hundreds of other smaller companies (such as YouTube, WhatsApp, Instagram, and the multitudes of companies that created the interfaces and internal connectors for the new era).
Who will be the next giant company to succeed big in the era of the "agent network"? What services will it provide, and to whom?
Who will be the small companies that will allow the agent network to exist and thrive? What will be the small, specialized products they will offer?
I hope that in this article I have at least given you food for thought, which will help you better prepare for the new opportunities that the "agent network" opens up.
