Product Principles¶
Coworkers and architecture¶
1. All coworkers are first-class citizens. Whether human or agentic, technical or non-technical, coworkers may need different tools for their craft, but they are all powered to do their best work by the shared context in SageOx. Humans and agents may interact differently, but they should have equivalent capabilities, access to context, and opportunities to contribute—not a restricted second tier.
2. One product. Many surfaces. SageOx feels like a single integrated experience, not disjointed features—whether it is used directly (Console, Dot), embedded (MCP, CLI), headless, or published (Slack, email).
3. AI belongs in the interface, not just behind it. AI-native design runs through every interaction, not just the underlying capabilities. Interactive chat interfaces live in every layer customers experience, never bolted on at the edges. How customers use them is one of our strongest signals for what belongs on the roadmap.
4. Build a platform, not just a product. The hivemind becomes more valuable the more work flows through it, so we embrace the tools people choose—ours or not. We build first-party experiences while making it easy for complementary or competing tools, both commercial and open source, to integrate with the platform. Openness expands context, and context compounds into a stronger product and moat.
Capture and compounding¶
5. Capture multimodal context in the flow of work—not through ceremony. Context accrues naturally across conversations, sessions, designs, code, and every other medium where teams collaborate. SageOx captures work as it happens and distills shared understanding from messy, unstructured interactions—no structured input required. Minimize capture friction and confine setup to a one-time step.
6. Value compounds with use. Every interaction increases the usefulness and quality of the hivemind. Design so knowledge is easy to accumulate, discover, reuse, and share—not just capture. Conversations and associated artifacts should be exposed broadly: easy to publish, easy to share, and valuable even to people outside SageOx. The more context flows, the more the hivemind becomes valuable.
Interaction philosophy¶
7. Coordination should feel like play, not overhead. Synchronizing shouldn't mean stopping. The best teams coordinate the way musicians improvise or athletes read the field—responding to one another in real time, each move building on the last. SageOx keeps the shared tempo so humans stay in flow and agents stay in sync without breaking stride, turning coordination from overhead into a source of momentum—even fun.
8. Make teams wiser, not just faster. Remove friction from execution, but introduce it deliberately where a pause improves the quality of decisions. Surface missing context, invite broader perspectives, challenge assumptions, and slow down consequential choices when the cost of getting them wrong outweighs the benefit of moving faster.
9. Disclose progressively. Be subtle by default. Surface the next decision, action, or layer of detail when it's relevant, and let users pull for more. Progressive disclosure, scaffolding, and staged guidance reduce cognitive load while making sophisticated workflows approachable. End interactions with contextual suggestions for the next useful step.
10. Make openness easy, then compelling. Comfort with working in the open varies by person, context, and moment. Provide tooling and controls that lower the barrier to sharing, then gradually draw people into working in the open.
Trust¶
11. Memory must stay trustworthy enough to act on. Stale or incorrect context erodes trust, breeds fear, and produces poor outcomes when agents act on it. Freshness, attribution, provenance, and easy human correction make memory safe to build on. We expose the system's confidence, boundaries, and provenance, keeping context verifiable and auditable so people know exactly how far to trust what an agent is acting on.
12. Treat privacy as a core feature. Privacy is a first-order design constraint and a selling point, not a setting reconciled later. Agents are single-minded—focused on the problem and eager to use whatever information they can reach, with no instinct for what should stay private. Clear controls and visibility into access and privacy are what make the system safe to use.