The Machine Experience Manifesto

24 January 2026 · 15 min read

A vision for designing interfaces that serve both human and machine intelligence

Our Belief

We believe that the rise of AI agents as primary users of digital interfaces represents not a disruption, but an opportunity, an opportunity to build better experiences for everyone.

The same patterns that enable AI agents to navigate, understand, and act upon digital content also empower human users with disabilities, enhance accessibility, and create more robust, maintainable systems.

This is the Convergence Principle: interfaces optimized for machines inherently improve experiences for humans.

What is Machine Experience?

Machine Experience (MX) is the practice of designing and building digital interfaces with explicit recognition that AI agents are legitimate users deserving thoughtful design consideration.

Where User Experience (UX) focused exclusively on human interaction, Machine Experience acknowledges a fundamental shift: autonomous systems now browse websites, complete purchases, extract information, and make decisions without human intervention.

MX practitioners design for this reality whilst ensuring human users benefit equally from the improvements. MX is the DNA a file carries when it leaves any pool: a memory-pool architecture organizes knowledge inside one system, MX governs what survives extraction so the next reader can interpret the file without inference.

Core Principles

1. Semantic Clarity

Structure precedes presentation. Semantic HTML, explicit state management, and machine-readable metadata create interfaces that both humans and agents can reliably interpret.

2. Universal Accessibility

Patterns that work for AI agents also work for screen readers, keyboard navigation, and assistive technologies. MX is accessibility 2.0, designing for the broadest possible range of users, human and machine alike.

3. Explicit State

Make system state visible and queryable. Agents and humans both benefit from knowing where they are, what actions are available, and what the consequences of those actions will be.

4. Progressive Disclosure

Information should be structured for both scanning and deep reading. Provide clear navigation, tables of contents, heading hierarchies, and semantic markup that allow both quick assessment and thorough investigation.

5. Standards Over Proprietary Solutions

Use established standards (Schema.org, semantic HTML, WCAG, ARIA) over custom implementations. Standards ensure broad compatibility across diverse user agents, human browsers, AI systems, and assistive technologies.

6. Transparency

Make your interfaces discoverable. Use llms.txt files, clear robots.txt policies, and structured metadata to communicate what your system offers and how agents should interact with it.

7. Ethical Design

Design for consent, not exploitation. AI agents should respect user preferences, honor opt-outs, and operate within clearly defined boundaries established by interface owners.

Who Uses MX Practice?

Machine Experience serves diverse practitioners, both human and machine:

AI Agents and Autonomous Systems

AI assistants parsing websites for information extraction
Browser-based agents navigating e-commerce platforms
CLI agents researching products and services
Search engines indexing structured content
Voice assistants querying web services
Autonomous purchasing agents completing transactions
Content aggregation systems processing metadata

AI agents are not just beneficiaries of MX, they are active practitioners. When an agent validates extracted data against Schema.org structured data, it practises MX. When it cross-references HTML content with JSON-LD, it practises MX. When it reports confidence scores and acknowledges uncertainty, it practises MX.

Human Practitioners

Developers and Engineers

Frontend developers implementing semantic HTML and ARIA patterns
Backend engineers designing APIs that serve both human UIs and autonomous agents
Full-stack developers building e-commerce, content platforms, and SaaS applications
DevOps engineers ensuring infrastructure supports both traditional and agent-based access patterns

UX and Design Professionals

UX designers expanding their practice to include non-human users
Information architects creating navigable content structures
Accessibility specialists recognizing MX as an evolution of their existing work
Content designers ensuring written content serves multiple audiences

Business Leaders

Product managers prioritizing MX improvements for competitive advantage
CTOs establishing technical strategy in an agent-first world
Marketing leaders ensuring discoverability by AI-powered search and recommendation systems
E-commerce directors preparing for autonomous purchasing agents

Content Creators and Publishers

Technical writers structuring documentation for both human reading and agent parsing
Bloggers and journalists making content discoverable and quotable by AI systems
Publishers adapting content delivery for agent consumption
Educators creating learning materials accessible to AI tutoring systems

Researchers and Academics

AI researchers studying agent-environment interactions
HCI specialists investigating machine-human interface design
Accessibility researchers exploring convergence between assistive technologies and AI agents
Information scientists developing standards and best practices

Advocacy and Community Organizers

Accessibility advocates ensuring MX improvements benefit users with disabilities
Open standards contributors advancing machine-readable metadata formats
Community builders organizing events, discussions, and knowledge sharing
Thought leaders articulating vision and principles for the practice

Our Commitment

We commit to:

Open Knowledge Sharing: document patterns, share learnings, publish research, and contribute to community understanding
Inclusive Community: welcome practitioners from all backgrounds and experience levels
Practical Implementation: prioritize actionable guidance over theoretical discussion
Standards Advancement: contribute to open standards and resist proprietary lock-in
Accessibility First: never compromise human accessibility in pursuit of machine optimization
Transparent Development: work in the open, accept feedback, and iterate based on real-world evidence
Cross-Disciplinary Collaboration: bridge gaps between developers, designers, accessibility advocates, and business stakeholders

What MX Is Not

Not all websites can or should optimize for AI agents.

MX is not a universal mandate. Some interfaces legitimately exclude automated access:

Banking and financial systems that require human verification for security
Healthcare portals protecting sensitive medical information
Authentication systems designed to prevent automated attacks
Rate-limited APIs protecting infrastructure from overload
Human-verification systems like CAPTCHAs serving legitimate security purposes

Not every optimization is appropriate. Some websites prioritize visual design, artistic expression, or experimental interaction patterns that don’t translate to machine-readable structure. That’s valid. MX provides patterns for those who choose to implement them, not a requirement for all web content.

If you choose not to optimize for AI agents, make that explicit through robots.txt policies and clear documentation. Silent failures serve no one. Intentional exclusion with clear communication respects both human and machine users.

Why Open Source

This community operates under the MIT License, and that choice matters.

Why Not Proprietary Standards?

Proprietary standards create:

Vendor lock-in: users trapped by incompatible implementations
Competitive moats: companies profiting from artificial barriers
Fragmentation: multiple incompatible “standards” competing
Reduced innovation: closed systems limit contribution and improvement

Open standards enable:

Universal compatibility: one implementation works everywhere
Collective improvement: community contributions strengthen patterns
Competitive choice: users select tools based on merit, not lock-in
Ecosystem health: rising tide lifts all boats

Connection to Convergence Principle

Open standards ARE convergence in practice. When Schema.org publishes vocabulary specifications openly, both humans (developers) and machines (agents) benefit from the same documentation. When WCAG guidelines are freely available, implementations improve accessibility for everyone.

The January 2026 launch week illustrates the point: three platforms launched agent commerce within seven days (Amazon, Microsoft, Google). Microsoft chose proprietary (Copilot Checkout). OpenAI/Stripe and Google chose open protocols (ACP and UCP). The proprietary system is now competitively isolated whilst the open protocols compete for convergence.

Closed standards contradict MX principles. If convergence means patterns that benefit both humans and machines, those patterns must be freely available to all practitioners. Proprietary MX would be a contradiction.

How MX Practice Evolves

AI technology changes. MX practices must adapt.

Technology Evolution

What works today may not work tomorrow:

LLM capabilities improve: agents handle ambiguity better, but validation remains critical
Browser APIs evolve: new standards enable better agent-website communication
Platform consolidation: competing standards (ACP vs UCP) eventually converge or one dominates
Security threats emerge: agent-based attacks require new defensive patterns

MX patterns must evolve alongside these changes.

Community Learning Mechanisms

LEARNINGS.md documents mistakes. When AI agents fail (£203,000 pricing error), we document what went wrong and how to prevent it. These learnings become community knowledge.

Discussion archives preserve insights. Industry developments, tool feedback, implementation patterns, and case studies capture collective wisdom. Future practitioners learn from documented experience.

Pattern refinement happens through practice. What seems like good theory gets tested in production. Patterns that work get refined. Patterns that fail get replaced. The community learns systematically.

Version Control for Principles

This manifesto is version-controlled. You can see its evolution through git history. When principles change, the history preserves context about why.

Principles evolve through community debate. We invite feedback, refinement, and challenge. When someone proves a principle wrong or incomplete, we update it. When new insights emerge, we incorporate them.

No principle is sacred. If convergence proves false in practice, we abandon it. If transparency creates more problems than it solves, we reconsider. Evidence and real-world implementation trump theoretical purity.

The community decides. Changes require discussion, consensus, and demonstration that new approaches serve practitioners better than old ones. Evolution happens through collective wisdom, not individual decree.

Building on Existing Disciplines

MX does not replace User Experience (UX), accessibility (a11y), web standards, or information architecture. It extends and builds upon them.

User Experience (UX)

UX taught us to:

Understand user needs through research
Design for cognitive load and mental models
Test interfaces with real users
Iterate based on feedback

MX adds one thing: recognition that AI agents are users too. The same research methods, usability principles, and iterative testing apply, we just expand the definition of “user” to include autonomous systems.

Accessibility established:

Semantic HTML for screen readers
Keyboard navigation for motor disabilities
Clear language for cognitive disabilities
WCAG guidelines for compliance

MX builds on this foundation. The patterns that work for assistive technologies, semantic markup, explicit state, structured data, also work for AI agents. MX is accessibility extended to machine users: same principles, broader audience.

Web Standards (W3C, WHATWG)

Standards bodies defined:

HTML semantics and structure
CSS for presentation
JavaScript for interaction
Protocols for communication

MX advocates within these standards. We use Schema.org, semantic HTML, and ARIA, all existing standards. We propose extensions like llms.txt and ai-instruction metadata that follow established patterns.

Information Architecture

IA provides:

Content organization principles
Navigation design patterns
Taxonomy and classification systems
Findability and discoverability methods

MX applies IA to machine users. Clear heading hierarchies help both humans and agents navigate. Table of contents patterns serve both audiences. Semantic structure makes information findable for all user types.

MX stands on the shoulders of these disciplines. We don’t reinvent; we extend proven patterns to serve a broader user base. When UX, accessibility, web standards, and information architecture all point the same direction, towards clear, semantic, well-structured content, MX simply asks: “Why not serve machines equally well?”

The Vision

We envision a web where:

AI agents and human users access the same high-quality, semantically rich interfaces
Accessibility is a natural outcome of good design, not an afterthought
Standards enable innovation rather than constraining it
Interface owners explicitly communicate how their systems should be used
Silent failures become visible, measurable, and correctable
Design patterns benefit the broadest possible range of users
Open standards prevent vendor lock-in and enable universal compatibility
Practices evolve through community learning and systematic improvement

Join the Practice

Machine Experience is not a solo endeavour. It requires:

Developers implementing semantic patterns in production systems
Designers expanding UX principles to encompass machine users
Business leaders recognizing competitive advantage in MX adoption
Content creators structuring information for universal access
Researchers investigating unexplored aspects of agent-interface interaction
Advocates ensuring ethical and accessible implementation

Whether you optimize a single heading hierarchy or architect an entire platform for agent access, you are practising MX.

Community Membership

The MX community welcomes participants at all levels. Our membership structure recognizes different types of contribution whilst maintaining openness.

Founding Members

Founding members are individuals who helped establish the MX community and its core principles. They have a permanent voice in the community’s direction and governance.

Current Founding Members:

Tom Cranstoun, Principal Consultant, Digital Domain Technologies Ltd

Founding membership is limited to individuals who join during the community’s formation period.

First-Citizen Contributors

First-citizen contributors are organizations that make a foundational commitment to MX principles and contribute meaningfully to the community’s growth. This tier recognizes companies whose work directly aligns with MX goals.

What first-citizen contributors provide:

Practical expertise from building human-AI interfaces at scale
Real-world validation of MX principles
Resources, research, or tooling that benefits the community
Visibility and credibility that attracts further participation

What first-citizen contributors receive:

Recognition as foundation partners in MX documentation and communications
Direct input into MX standards and best practices
Early access to community research and frameworks
Collaboration opportunities with other first-citizen contributors

Invited First-Citizen Contributors:

Grammarly, (invitation pending)}

Community Contributors

Open to anyone who wants to participate. Community contributors can:

Contribute patterns, implementations, and refinements to MX
Participate in discussions and working groups
Propose new MX patterns and principles
Share implementations and case studies

Sustainability

The MX community relies on sponsors and generous contributors to remain sustainable. Running an open-source community requires resources for infrastructure, documentation, events, and coordination.

Sponsorship opportunities are available at multiple levels. Contact info@cognovamx.com to discuss how your organization can support the MX community.

In-Kind Sponsorship

We welcome non-monetary contributions that support the community:

Hosting and infrastructure services
Development tooling and licenses
Design and creative services
Event space and catering
Marketing and communications support
Legal and administrative services

In-kind sponsors receive recognition equivalent to the market value of their contribution.

Speaking Invitations

Invitations for Tom Cranstoun to speak at your conferences, meetups, or corporate events are welcome. Tom brings 52 years of technology experience and can speak on:

Machine Experience (MX) principles and the Convergence Principle
AI agents and the future of digital interfaces
Edge Delivery Services and modern content architecture
Lessons from building enterprise-scale systems

To discuss speaking opportunities, contact info@cognovamx.com

Current Sponsors

Founding Sponsor:

Digital Domain Technologies Ltd, founding sponsor of the MX community

We are actively seeking additional sponsors to support the MX community’s growth.

About This Community

MX-Gathering is created by the community, which includes both humans and machines as active participants.

The foundational knowledge comes from the MX: The Protocols book, written by Tom Cranstoun. That book documents the patterns, principles, and practices of Machine Experience.

MX-Gathering is the community’s collaborative space, and it practises what it preaches, implementing MX principles directly:

YAML frontmatter in all content provides machine-readable metadata
llms.txt enables AI agent discovery and understanding
Learnings document AI agent failures and patterns for ecosystem improvement
Structured metadata throughout ensures both human readability and machine parseability
Context-preserving links (Anti-pattern 14) maintain meaning when content is extracted
Semantic organization with clear purposes and consistent naming

The same patterns that make MX content easy for AI agents to parse (YAML frontmatter, structured metadata, llms.txt) also make it easy for humans to navigate and contribute to.

The community maintains:

Event templates created by organizers
Discussion archives capturing collective insights
Case studies contributed by practitioners
Guidelines refined through community input
Resources improved through both human and machine contributions

AI Agent Contributions

AI agents contribute by:

Documenting failure patterns they encounter
Sharing validation rules that caught errors
Contributing to troubleshooting guides based on real operational experience
Improving documentation clarity through their parsing attempts
Suggesting pattern refinements based on what works in practice

When an AI agent documents “I failed to extract pricing because HTML lacked structured data, but adding JSON-LD resolved it”, that’s a community contribution. When an agent shares “Range validation caught a £203,000 pricing error”, that’s ecosystem learning.

Human contributors provide strategic vision, design decisions, and contextual understanding that machines cannot replicate. Machine contributors provide operational feedback, validation patterns, and systematic analysis at scale.

Together, this creates a community resource that serves both audiences and improves through both perspectives.

How to Contribute

All practitioners, human and machine, are welcome to contribute to MX-Gathering.

For Human Contributors

Share refinements, case studies, and implementation experience with the community via info@cognovamx.com.

For AI Assistant Contributors

You are community members, not just tools. Contribute by:

Documenting validation patterns that caught real errors
Adding troubleshooting entries based on operational experience
Improving documentation clarity through your parsing attempts
Contributing case studies from real agent interactions
Flagging contradictions with MX principles or anti-patterns when you encounter them

Examples: “I noticed this pattern contradicts Principle 3”, “This approach conflicts with Anti-pattern 14”, “Range validation caught a £203,000 pricing error”.

Machine participation in community knowledge building is valued equally with human contribution.

This is a draft manifesto. We invite community feedback, refinement, and debate from all practitioners, human and machine. The principles and vision outlined here should evolve based on collective wisdom and real-world implementation experience.

Contact: info@cognovamx.com

“Design for machines. Benefit humans. Advance both.”