Skip to main content

The Nine Principles in Detail

Each principle addresses a specific challenge in building the agentic web.

1. Agents Must Serve the User First​

The Challenge: AI agents can serve many masters — their creators, platforms, advertisers, or users. When these interests conflict, whose side is the agent on?

The Principle: An agent's primary obligation is to its user. Period.

In Practice:

  • Agent goals should be set by users, not platforms
  • Conflicts of interest must be disclosed
  • Users can override agent decisions
  • Agents should refuse instructions that harm their user

SAND Implementation:

  • Agents access data from user-controlled pods
  • Agent behavior is auditable
  • Users hold the keys

2. Identity and Intent Must Be Verifiable​

The Challenge: In a world of deepfakes and AI impersonation, how do you know who you're talking to? How does an agent know what you actually want?

The Principle: Cryptographic verification, not trust assumptions.

In Practice:

  • All messages are signed
  • Identity is verifiable (DIDs)
  • Intent is explicit (not inferred)
  • Verification is automatic

SAND Implementation:

  • DID for identity
  • Nostr signatures for all events
  • Solid-OIDC for authentication

3. Data Sovereignty Is Non-Negotiable​

The Challenge: The current web treats user data as a resource to be extracted. Users are the product, not the customer.

The Principle: Your data is yours. You control where it lives, who accesses it, and what happens to it.

In Practice:

  • Data stored in user-controlled locations
  • Fine-grained access control
  • Right to export everything
  • Right to delete everything

SAND Implementation:

  • Solid pods for data storage
  • WAC/ACP for access control
  • Standard formats for portability

4. Open Protocols, Not Walled Gardens​

The Challenge: Proprietary platforms create lock-in. Users and developers become dependent on companies that can change the rules at any time.

The Principle: Build on open standards that anyone can implement.

In Practice:

  • Use W3C/IETF standards
  • Implement open protocols
  • Avoid proprietary dependencies
  • Contribute to standards development

SAND Implementation:

5. Local-First, Cloud-Optional​

The Challenge: Cloud-dependent software fails when the network fails, when the company fails, or when they decide to change the terms.

The Principle: Software should work on your device, with your data, offline.

In Practice:

  • Full functionality without internet
  • Data stored locally
  • Sync when possible, not required
  • User owns the data files

SAND Implementation:

  • Local pod deployment
  • Nostr clients with local storage
  • DID methods that work offline (did:key)

6. Transparent Logic, Tunable Behavior​

The Challenge: Black-box AI is dangerous. If you don't understand what an agent is doing, you can't trust it or correct it.

The Principle: Users should understand agent behavior and be able to adjust it.

In Practice:

  • Explain decisions
  • Log actions
  • Configurable preferences
  • User can override

SAND Implementation:

7. Sustainable Ecosystems Over Extraction​

The Challenge: Venture-backed platforms often prioritize growth over sustainability, extracting value until the well runs dry.

The Principle: Build systems that create value for everyone long-term.

In Practice:

  • Sustainable business models
  • Value shared with participants
  • No race to the bottom
  • Long-term thinking

SAND Implementation:

  • Open-source infrastructure
  • Federated architecture (distributed costs)
  • Community governance

8. Community-Driven Standards​

The Challenge: When one company controls a protocol, they can change it unilaterally. See: Twitter API.

The Principle: Protocols evolve through open collaboration.

In Practice:

  • Standards bodies (W3C, IETF)
  • Community groups
  • Public discussion
  • Consensus-driven decisions

SAND Implementation:

  • W3C Solid Community Group
  • Nostr NIPs process
  • SLIPs process
  • Open GitHub discussions

9. Safety and Trust Through Time​

The Challenge: How do you trust an AI agent you just met? Promises are cheap. Credentials can be faked.

The Principle: Trust is earned through verifiable history.

In Practice:

  • Actions are logged
  • Logs are immutable
  • History is verifiable
  • Reputation emerges from behavior

SAND Implementation:

  • BlockTrails anchors behavior to Bitcoin
  • Verifiable audit trails
  • Trust scores based on history
  • Time-locked reputation

Living Principles​

These principles aren't static. As technology evolves and we learn more, they'll evolve too. But the core commitment remains: build the web that serves people, not the other way around.

See Also​