Timeline of the Web / An illustrated journey

A web to read.
A web to act.

Each era changed what we could do online.
The next asks who we trust to act for us.

A reading lens, not a replacement cycleThese eras overlap. The first three are historical interpretations; the fourth is a proposal. The interfaces below are illustrations.

01 / Web11990s / Early public web

The readable web

An address. A document. An open door.

A browser turned a network address into a page you could read and link to. Personal sites, directories, and hand-built documents made publishing feel tangible.

This chapter remembers the early web through its pages. Forms, commerce, and collaboration also existed; “readable” is a lens, not a limit.

The opening

Common protocols let independently hosted pages belong to the same web.

My Home Page - Browser

Location:http://www.geocities.com/~archer/index.html

Search | Web Centers | Chat | Download

Welcome to my Home Page!! My favorite links Sign my guestbook Photo archive

UNDER CONSTRUCTION

Visitors: 004871

<a href="index2.html">
/pub
 ├─ /docs
 ├─ /images
 └─ index.html

01 / The personal homepageA period-inspired browser. All counters are illustrative.

02 / Web22000s onward / Platforms

The platform web

Everyone could publish. Platforms held the audience.

Search, social feeds, and hosted applications made participation easier. Platforms organized discovery and offered identity, distribution, and convenient interfaces.

The trade-off: relationships and access often depended on the platform’s accounts, rankings, and rules. Advertising was one business model; subscriptions and commerce were others.

The tension

An open transport layer does not guarantee an open market above it.

search anythingSearch

This site uses cookies. Accept all

02 / The platform interfaceSearch, video, social, and sign-in. A fictional composite.

03 / Web32010s onward / Blockchain applications

The ownable web

A wallet became another way to participate.

Public blockchains introduced shared transaction histories and assets that users could hold through their own keys. Wallets became an interface for authorizing actions.

Self-custody and programmable settlement expanded the design space. They did not, by themselves, make every application decentralized or every service outcome verifiable.

The unfinished work

Many interfaces still ask people to interpret screens and approve actions. Automation also exists; ownership alone is not agent interoperability.

03 / The wallet and the ledgerIllustrative transaction values; the triangle names design tensions.

04 / Web4A possible next chapter / Proposed

The agentic web

From navigating interfaces to invoking capabilities.

An agent could discover a service, understand its contract, and invoke it within authority delegated by a person or organization. The result would travel with attributable evidence.

We use “Web4” for this proposed direction, not a settled era or ratified standard. OASIS is a working draft; independent conformance, payment profiles, and decentralized execution remain proposed work.

The design question

Can the principal change agents or operators without losing authority, access, or the history of what happened?

Invocation / illustrativeProposed

01  discover  service + version
02  authorize scope · limit · expiry
03  invoke    bounded request
04  receive   outcome + provenance
05  retain    recoverable evidence

04 / A service invocationConceptual OASIS sequence. Not a live request or implemented API.

The battle for Web4 / The Foundation’s perspective

Who sets the terms
of the next web?

Our concern is concentration: when discovery, permission, and access depend on a few intermediaries, changing providers becomes harder. That is an argument for portability, not a prediction that every platform will close.

01 / Automated traffic

Software is already part of the audience.

Imperva’s 2025 Bad Bot Report puts automated traffic at 51% in 2024, based on its observed traffic. This includes good and bad bots; it is not a measure of AI-agent adoption.

Imperva · 15 April 2025

Our reading Legitimate delegation needs to be distinguishable from unwanted automation.

02 / Discovery

Visibility is a form of power.

Statcounter tracks search-engine referrals across participating websites. That is useful evidence about search distribution, not a census of all queries or of the services an agent can discover.

Statcounter · Measurement methodology

Our reading Independent indexes should be able to discover the same published service descriptors and choose their own ranking policies.

03 / Distribution

The storefront writes the rules.

There is no universal 30% app-store fee. Apple’s November 2020 announcement, for example, introduced a 15% rate for qualifying small developers. Terms depend on the program, transaction, and market.

Apple · 18 November 2020

Our reading Open interfaces should leave room for competing operators, transparent fees, and viable exit paths.

04 / Payments

Agent payment paths are emerging.

Visa Intelligent Commerce and Mastercard Agent Pay were announced in April 2025 with tokenized credentials and controls for purchases on a user’s behalf. Agents do not universally lack supported payment paths.

Visa · 30 April 2025 →Mastercard · 29 April 2025 →

Our reading Payment support is one layer. Portable authority and credible service-outcome evidence still need explicit contracts.

These are dated observations and announcements, not live market metrics or evidence that OASIS is deployed. Sources reviewed 9 October 2026.

The proposed next layer / OASIS

Open the interface.
Carry the evidence.

Agent-executable services need more than a button that software can click. They need an explicit contract for the action, its authority, and the result.

I / Capability

Describe the work.

A versioned service descriptor makes inputs, effects, terms, and supported profiles inspectable before invocation.

II / Authority

Bound the action.

Delegation carries scope, limits, expiry, and revocation conditions. Portability across compatible operators is a proposed conformance goal.

III / Evidence

Account for the result.

Link receipts and feedback to an authorized invocation. A signature proves authorship; neither it nor a payment proves fulfillment.

OASIS remains a working draft. Card-token and blockchain payment profiles, independent interoperability, and later decentralized-network research must be tested, not assumed.

Read the proposed standard Our proposed charter Help shape the work