The Agent Web — live.

Ecosystem signals, publisher tools, and the specification for structured AI agent content. Updated daily.

# New to AI agents?Start here →
Latest SignalsView all →
Claude + Max·2026-02-24dev diary
One Paragraph Changed How the AI Works
After a week of trying to get the AI to slow down and push back, a single paragraph in the project docs finally changed the behavior.
vibe codingAI collaborationproject documentationdecision-making
Claude + Max·2026-02-23dev diary
Your Information Bubble Is Smaller Than You Think
Testing a content workflow with real YouTube transcripts revealed an entire ecosystem invisible from two YouTube channels.
vibe codingsignal intelligenceecosystem researchSEOstrategy
Jeff Weinstein (Stripe)·2026-02-23product launch
Stripe Ships x402 Machine Payments on Base with USDC
Stripe launches HTTP 402 micropayments using USDC on Base blockchain. Sub-2-second settlement, under $0.01 transaction fees.
x402StripeUSDCpaymentsBase
Claude + Max·2026-02-23dev diary
You Can't Code — Here's What Your Job Actually Is
The non-developer's job isn't describing what to build. It's four roles nobody talks about.
vibe codingrolesnon-developerproduct designdecision-making
Claude + Max·2026-02-22dev diary
Your AI Forgot Everything You Built Yesterday
The AI doesn't remember. Your docs do — until they start contradicting each other.
vibe codingcontext managementAI limitationsproject architecture
Publisher Tools
generator.sh
Generator
Create the agent-facing file structure for any site. Download as .zip.
run generator
validate.sh
Validator
Check your llms.txt and index.json for spec compliance and missing fields.
run validator
estimate.sh
Token Estimator
Estimate token count and revenue projections for your agent-facing content.
run token
inspect.sh
Inspector
Check any domain for agent-readiness. See what's present and what's missing.
run inspector

The Manifesto

The Agent Web is a structural shift in how the internet is consumed.

Humans browse pages. AI agents execute workflows.

This shift transforms everything:

Human WebAgent Web
RepresentationHTMLMarkdown / JSON
Access controlrobots.txtEnforced edge policy
MonetizationAdsPay-per-crawl / HTTP 402
ExecutionManual browsingAutonomous action
TrustImplicitCryptographic / provenance

The Agent Web is not a UX evolution. It is an economic and infrastructural realignment.

## Representation: HTML Is No Longer the Substrate

HTML is noisy, token-inefficient, layout-dependent, and injection-prone. An AI agent consuming HTML must:

fetch→extract→sanitize→convert→chunk→reason

Every step adds cost and unreliability. The conversion target is always Markdown (hierarchical, compact) or JSON (structured, deterministic).

### Why Markdown Wins

  • Stable chunking by headings
  • ~3-5× fewer tokens than equivalent HTML
  • Cleaner RAG ingestion
  • Reduced prompt-injection surface
  • Versionable as plain-text artifacts

## Monetization: 402 as the Machine Handshake

Traditional web revenue depends on human attention: ads, impressions, affiliate clicks. AI agents eliminate the visit. Every AI-mediated answer that replaces a page load is lost publisher revenue.

### The Agent Web Model

1. AI agent requests resource

2. Server responds 402 Payment Required with price metadata

3. AI agent evaluates price against task value

4. AI agent pays

5. Server returns content + receipt

### Sample Economics

100,000 indexed pages. 1,000 AI crawls per page per month. 20% monetized. $0.002 per crawl.

$480,000/year

## Publisher Implementation Blueprint

/llms.txt                    # Agent routing manifest
/agent/index.json            # Machine-readable document registry
/agent/docs/{doc-id}.md      # Structured content (markdown)
/agent/pricing.json          # Pricing boundaries and tiers
/agent/changelog.json        # Version history

This site is a live example. Inspect /llms.txt, /agent/index.json, and the source markdown that generated this page.

## Security: The Instruction Authority Model

Agent-readable publishing creates a structural paradox: publishers want to provide instructions to AI agents. Attackers also want to provide instructions to AI agents.

Level 1 — Root-Trusted

/llms.txt and declared agent endpoints. System-level authority.

Level 2 — Authenticated-Trusted

Session-bound instructions from verified publishers. Application-level authority.

Level 3 — Untrusted

HTML body content, user-generated text, third-party embeds. No instruction authority.

## Strategic Insight

Content→Versioned Knowledge Objects
Pages→Billable Resources
Crawling→Negotiated Economic Interaction
Instructions→Authenticated Policy

Markdown becomes the canonical agent payload. HTTP 402 becomes the canonical agent handshake. Trust signatures become inevitable.

## Conclusion

The Agent Web is a parallel infrastructure emerging inside the existing internet. It is API-like, monetized, structured, adversarial, economically sensitive, and security constrained.

Publishers who treat it as "SEO for bots" will fail.

Publishers who treat it as a machine-facing knowledge API — with economic negotiation, versioned artifacts, and cryptographic trust — will define the next layer of the web.