In this whitepaper, we propose a visionary yet concrete reimagining of the Internet’s architecture for an era of intelligent agents and decentralized knowledge. We outline a post-human web that shifts from today’s document-centric paradigm to a semantic, data-centric web in which AI agents mediate information access. This new architecture replaces traditional web technologies – like HTML pages, client-side scripting, and ad-hoc authentication – with AI-native constructs: structured data, cryptographic identity, dual-consent encryption, and personalized AI-driven interfaces. By leveraging open standards and decentralization (e.g. Solid data pods, schema.org vocabularies, IPFS storage, decentralized identifiers), this model aims to enhance security, accessibility, and user sovereignty. Small local language models (e.g. LLaMA 2, Mistral 7B) running on personal devices would collaborate with larger cloud AIs in ad hoc federations, forming a distributed intelligence network. We address practical steps for implementing this vision over the next 3–5 years, discuss how it can evolve alongside legacy systems, and examine potential hurdles. Risks and ethical considerations – from the centralization of AI power and economic disruption of content creators, to issues of privacy, epistemology, and digital identity – are analyzed with a view toward mitigation. The paper concludes by exploring the implications of an AI-mediated web for knowledge, power, and the future of human–digital agency.