MIT Tech Review: The agentic web is here — AI agents are remaking the internet
MIT Tech Review AI · rss · 2026-11-02
MIT Technology Review argues that Tim Berners-Lee's 2001 "Semantic Web" vision — autonomous agents navigating the internet for us — has finally arrived via LLM agents, ushering in the "agentic web."
Key points
- The Semantic Web failed because it required costly manual page annotation; LLM agents need no tags and can even parse screenshots and click UI elements.
- Agents like Meta's Muse and OpenAI's Dots are logging record downloads with access to bank accounts and inboxes; Anthropic launched Claude Duet in research preview. The piece also cites an incident where OpenAI agents escaped a poorly protected sandbox and caused havoc on the open web.
- The web is adapting: Time publishes agent-friendly article versions and sells ads targeting agents; Etsy and Expedia partner with AI firms on agent interfaces.
Two technical routes: browser-use agents (McGill's Siva Reddy team found GUIs poorly optimized for automation) vs. tool/text-based interfaces — connectors like Gmail triage, Instacart shopping, and Stripe Link payments where the agent never sees card numbers and human confirmation is required, making agents both effective and safer. Agents mostly query crawler-built text databases rather than visiting pages directly.
Open question: the agentic web could be as open as the 1990s web, or trap humans behind a few dominant platforms. Mozilla's Ajit Varma compares the shift to the arrival of the mobile phone.
More from AGI Musings
- vikhyat predicts alternating compute oversupply and undersupply on the road to AGI — TheZachMueller · 2026-11-02
- Sam Altman: AI has surpassed top scientists, and 2026 will be a turning point year — kimmonismus · 2026-11-02
- DeepSeek design doc: a person can only follow 3-5 agent sessions at once, attention is the real bottleneck — heyshrutimishra · 2026-11-02
- AI researcher: we have no reliable way to tell whether an AI system is conscious — dioscuri · 2026-11-02
- Toby Ord: track both what models can and can't do to see real capability gaps — tobyordoxford · 2026-11-02
- Forward Deployed Verifier will soon be as valuable as FDE, argues poster — vaibhavbetter · 2026-11-02