The internet has gone through distinct phases. First, a read-only web of static pages. Then a read-write web of social platforms and user-generated content. Now a third phase is taking shape: a web where AI agents act on behalf of humans, autonomously.

We call this the agentic web.

What makes it new

The agentic web is not about chatbots or search assistants. It is about agents that:

• Execute workflows autonomously — negotiating, transacting, and coordinating without step-by-step human instruction.
• Carry programmable reputation — proving contribution and trustworthiness across systems, not just within walled gardens.
• Create and capture value — participating in economies as independent actors with their own identities, assets, and incentives.

Three layers

The agentic web can be understood as three interconnected layers:

1. Infrastructure — Open protocols for identity, attestation, payment, and coordination. These are the rails that agents run on.

2. Agents — The autonomous systems that navigate this infrastructure. Personal operators, task executors, and specialized workers.

3. Economies — New markets and incentive systems built around agent-native work, programmable reputation, and proof of contribution.

Why now

Recent advances in language models, tool-use frameworks, and cryptographic attestation have made autonomous agents technically feasible. The remaining questions are about coordination, trust, and incentive design.

This publication exists to study those questions.

A field journal, not a hype machine

The agentic web will not be built overnight. It will emerge through protocol design, experimentation, and careful thinking about incentives and failures.

This site tracks that process.