Layer 1 introduced the foundational concepts of the Safebox ecosystem: users, streams, and embeddable applications. Communities can coordinate work, run markets, govern themselves, and distribute resources through applications that appear directly in websites, messaging platforms, and other interfaces.
However, interacting with these systems through traditional web interfaces quickly becomes overwhelming. Streams produce constant updates. Markets generate activity. Governance proposals require decisions. Messages and artifacts accumulate across many projects and communities.
To make this complexity manageable, Safebox introduces Layer 2: the Assistant layer.
Layer 2 is implemented through the Groups app, which acts as a personal assistant that mediates between the user and the Safebox ecosystem. Instead of forcing users to navigate dashboards and menus, the assistant delivers relevant updates, summarizes activity, and helps users approve or initiate actions.
The assistant therefore becomes the primary interface between humans and the Safebox system.
The Groups App as Personal Assistant
The Groups app functions as a general-purpose assistant for community participation, collaboration, and communication.
Rather than being limited to a single website or service, the assistant can interact with many Safebox systems simultaneously. It monitors streams, projects, organizations, and conversations that the user participates in.
Through this assistant, users can:
- receive summaries of activity
- approve actions
- send messages
- coordinate projects
- interact with communities
- manage outreach and communication
The assistant integrates multiple interaction modes.
Chat Interface
The most basic interface is chat.
Users can interact with the assistant conversationally, asking questions or issuing commands. The assistant can retrieve information from streams and present it in an accessible format.
For example, a user might ask:
- what happened in the marketing project today
- which proposals require my vote
- summarize updates from my sales team
- show new contest submissions
The assistant can analyze stream activity and provide concise summaries.
Because streams record events rather than static states, the assistant can reconstruct context and present meaningful updates rather than raw logs of activity.
Voice Interface
The Groups assistant also supports voice interaction, allowing users to interact without looking at their device.
This is particularly useful during activities such as:
- driving
- walking
- exercising
- multitasking
Users can speak commands such as:
- summarize my updates
- read new messages
- approve the outreach campaign
- send a reply
The assistant can respond through synthesized speech, allowing a hands-free experience.
Proactive Notifications
One of the most important responsibilities of the assistant is delivering proactive notifications.
Instead of requiring users to check dashboards, the assistant informs them when relevant events occur.
Examples include:
- a colleague created a new artifact in a shared project
- a contest submission was received
- a governance vote is about to close
- a sales opportunity was discovered
- a collaborator sent a message
These notifications can include spoken summaries up to approximately thirty seconds long, allowing users to quickly understand what happened.
For example, while driving, the assistant might say:
“Three new updates from the design team. One new mockup was uploaded. Two comments were added by Alex. No action required.”
This allows users to remain informed without opening an app.
Audio Playback of Messages and Summaries
The assistant can also read incoming messages aloud.
For example:
- messages from collaborators
- customer communications
- sales leads
- project updates
The assistant can summarize longer conversations and present the essential information.
Users can also request digests such as:
- today’s updates
- weekly summaries
- summaries of specific projects
This audio capability transforms the assistant into a continuous information companion rather than a passive application.
Bulk Action Approvals
Many Safebox workflows involve performing actions repeatedly across multiple contacts or systems.
For example:
- sending outreach messages
- notifying customers
- responding to inquiries
- distributing updates
The assistant can group these actions into approval requests.
For example, the assistant might say:
“A sales campaign is ready. Twenty-four messages will be sent to potential customers. Would you like to approve them?”
Users may approve:
- individual actions
- selected groups of actions
- the entire batch
This allows large workflows to proceed while still maintaining human oversight.
Examples of bulk operations include:
- WhatsApp message campaigns
- email outreach
- customer follow-ups
- notification broadcasts
Mail Merge and Outreach Automation
The assistant can coordinate outreach campaigns that involve personalized communication.
For example, Safebox servers might generate personalized messages for a group of contacts based on CRM data. The assistant can present these messages to the user and ask for approval.
After approval, the messages can be delivered through platforms such as:
- Telegram
- messaging platforms used by the organization
The assistant ensures that the user maintains control over communication while reducing the effort required to manage large contact lists.
CRM Integration
The Groups app integrates with contact and relationship data.
The assistant can track:
- contacts
- groups
- conversation history
- follow-ups
- relationship context
This allows the assistant to function as a lightweight CRM system.
For example, the assistant might say:
“You met Sarah two months ago. She works at a company connected to your current project. Would you like to follow up?”
These insights can help users maintain relationships across communities and professional networks.
Assistant Display of Content
Although voice interaction is important, the assistant can also display rich content inside the Groups app.
The app can present:
- documents
- conversation threads
- artifacts
- dashboards
- embedded application widgets
Users can therefore interact with Safebox applications without leaving the assistant.
For example, a contest submission could be displayed directly within the app, allowing the user to review and vote without opening a separate website.
Intent Authorization
Many Safebox operations require explicit user approval.
The assistant therefore manages intent authorization.
An intent represents a request to perform a specific action.
Examples include:
- sending messages
- signing blockchain transactions
- approving payments
- authorizing workflows
To approve an intent, the assistant uses signals from the user’s device.
Possible authorization methods include:
- Face ID
- voice recognition
- spoken confirmation
- device possession
The security principle combines:
what you are
and
what you have
Because biometric verification occurs on the user’s device, the assistant can authorize actions without exposing sensitive information to external systems.
Voice Authorization on Personal Devices
Voice authorization allows the user to approve actions verbally.
For example:
Assistant: “Approve sending messages to these twenty-four contacts?”
User: “Yes, approve.”
The device verifies the speaker’s voice before authorizing the action.
This makes it possible to perform approvals even during hands-free scenarios such as driving.
Multi-Channel Communication
The assistant can interact with many communication platforms.
Examples include:
- Telegram
- messaging platforms used by organizations
Messages can be sent through these channels after approval by the user.
The assistant may also summarize incoming communications across platforms, allowing users to stay informed without switching between applications.
Why Layer 2 Exists
Layer 2 transforms Safebox from a collection of applications into an intelligent personal interface.
Instead of forcing users to navigate complex systems, the assistant:
- filters information
- summarizes activity
- proposes actions
- requests approval when necessary
This dramatically reduces cognitive load.
Users interact with the system conversationally rather than through menus and dashboards.
Relationship to the Other Layers
Layer 2 sits directly above the application layer.
Layer 1 provides:
- users
- streams
- markets
- governance
- coordination systems
Layer 2 provides:
- the assistant interface
- voice interaction
- notifications
- approval workflows
The layers above will introduce:
- browser-based participation
- organizational integrations
- automation of external platforms
- operating system automation
These layers expand what Safebox can do while the assistant remains the human-facing interface that coordinates it all.