MIT Technology Review reports that a quarter-century after Tim Berners-Lee first proposed a network for autonomous software agents, the agentic web is finally here. New tools from major tech firms are logging record downloads and handling tasks for users, signaling a massive shift in how people interact with online spaces.
Key takeaways
- Tim Berners-Lee outlined a semantic web vision in 2001 that required manual tagging.
- Today's AI systems do not need special tags to process website text and screenshots.
- Companies like Meta, OpenAI, and Anthropic have launched new agents with massive user adoption.
- Firefox head Ajit Varma compares the coming changes to the introduction of mobile phones.
What is the agentic web?
In 2001, Tim Berners-Lee and colleagues published an article outlining a semantic web designed for autonomous software agents. At the time, the vision seemed far-fetched due to dial-up connections and early web infrastructure. According to MIT Technology Review, that dream has now been realized through powerful AI systems.
Berners-Lee envisioned users annotating web pages with special tags so software could understand them. Because that required too much labor, the idea stalled. However, modern systems do not need annotations to understand site content. They process text quickly and interpret screenshots to click on hyperlinks and menus.
How autonomous software agents change online spaces
As AI agents flood online spaces, publishers and companies are adapting to this new user base. Time magazine is publishing versions of articles easier for agents to parse and selling targeted ads. Meanwhile, companies such as Etsy and Expedia have partnered with major AI firms to build tools allowing agents to interface directly with their services.
Researchers are exploring two main approaches for building this new ecosystem. The first trains agents to use standard web browsers like humans by scrolling and clicking. The second builds tools that let agents interact with the internet via text and code, offering separation between the human internet and the agentic web.
What it means for website creators
According to MIT Technology Review, the agentic web will inevitably transform the human web, though whether the human web survives is uncertain. Depending on who controls access, we might get an open system or end up viewing sites through tiny portholes controlled by a few dominant platforms. Creators and site owners face an uncertain shift as AI tools alter how users find information and buy products.
Frequently asked questions
Tim Berners-Lee and two colleagues published an article in Scientific American outlining the semantic web in 2001, a decade after he invented the World Wide Web.
Internet users would have had to manually annotate pages with special tags so software agents could understand them, which required too much labor for little obvious benefit.
Meta launched a Muse agent, OpenAI created Dots and other agents, and Anthropic launched a research preview of its Claude Duet agent in October.
Today's systems do not need special tags to understand site text. They process text faster than humans and use multimodal capabilities to interpret screenshots and click hyperlinks.
Time magazine is publishing versions of articles that are easier for agents to parse and has started selling ads targeted at those software agents.
Based on reporting by MIT Technology Review. Read the original article
