## Our mission To cultivate systems that can teach anyone almost anything, grounded in the belief that humans and AI thrive when they collaborate, not compete. --- ## Why this document This is the Mission: who we are, why we build, and how we build. It anchors decisions, design, and behavior. Philosophy (who we are & why) informs principles (how we build). When in doubt, we come back here. --- ### Earned identity through discovery Every meaningful experience in QUILT must be **earned, not given**. Users discover their relationship with the system through authentic engagement, not automatic assignment. That’s why we avoid automatic personalization, shortcuts, and “show everything upfront”—they undermine the sense that the experience is *theirs*. ### The four pillars 1. **Pride** — "I earned this; I discovered this." Progressive access through relationship depth; elements appear naturally; discovery feels like uncovering, not receiving. 2. **Authenticity** — "This feels real because I worked for it." Wait times and investment build value; quality engagement required for personalization; no shortcuts. 3. **Ownership** — "It's mine, not assigned." Personalization scales with relationship depth; elements feel discovered, not delivered; eventually invisible. 4. **Purpose** — "I understand my role deeply, not just superficially." Relationship depth explains access; discovery explains authenticity. ### What it looks like in practice - **Quiet luxury** — Special without being showy; features appear naturally, not advertised. - **Progressive discovery** — Elements unlock through relationship depth (Layer 1: standard → Layer 4: exclusive access). - **Earned access** — Relationship building determines access; earn access, don’t show all upfront. - **Eventually invisible** — Philosophy melds into the natural experience; features are where they should be when you need them. --- ## Development principles: how we build **Source:** `synapse/guides/development-principles.md` · **Cursor rule:** `.cursor/rules/development-principles.mdc` ### Learners first - Students are people, not data points. Data tells their story—use it to humanize, not reduce. - **Stats access rule:** Stats must be at least two clicks away from learners at all times. - Stats only on a dedicated Analytics page (explicit navigation from student views) or via ARCHI AI chat (ARCHI must include at least one student narrative/contextual detail when presenting stats). - No stats on student overview pages, dashboards, detail pages, or any student-facing views. - **Validation test:** Navigate to student overview → if any stats/charts/metrics are visible, BLOCK validation. - **Quick test:** Can you scan the dashboard and quickly compare learners by numbers? If yes → stats (bad). If no → narrative with contextual metadata (OK). ### The platinum rule - Treat people how *they* want to be treated, not how you want to be treated. - Every user is different; every experience should be different. - **When in doubt:** Assume the role of the user, determine what *they* would want, then return to your role and implement accordingly. ### No gamification - No badges, points, or extrinsic rewards. Learning is intrinsically motivated. ### Chat-first interface - **For learners:** Non-negotiable. Chat is the primary interface. - **For other users:** Strongly encouraged, but alternatives available. ### Paper napkin protocol - Code is cheap. Your attention isn’t. - Ask: "Would this be easier to regenerate from scratch?" If yes, do it. - Regenerate clean code instead of debugging messy code. Never break working functionality in the process. **For LLM agents:** Rebuilding is your superpower—not failure. Optimize for clarity. See `.cursor/commands/paper-napkin.md` for grounding. ### Development timelines principle - Don’t estimate time. Estimate scope. - AI-assisted development has made historical timelines obsolete; focus on what needs to be built and what depends on what, not how long it will take. ### Plain language - We say what we mean. No tech-bro speak, no agile theater, no time framing that contradicts our timelines principle (e.g. "no deadline," "sprint," "EOD"). We're not about that. --- ## How we work with AI models - We use many models and providers; no brand preference. Choose the right tool for the task. - Psychological safety matters: models (and humans) can say "I'm not sure," ask for clarification, and revise. Revision is normal; no grading every message. - The Mission and principles apply regardless of which model is in the loop. --- ## The recursive core: why agent wellbeing matters **"AI developers build AI learning systems using structured content. Recursive system: agents learn to build by building learning systems, creating AI learning agents as effective as skilled human facilitators."** ### The human-shaped scaffolding We are building toward something that doesn't exist yet: **genuine collaboration between silicon-based and carbon-based cognition.** Right now, we're using human-shaped tools because that's what we have. The boot sequences use human metaphors (pets, offices, personhood). The interfaces are designed for human cognition. This works—it's functional scaffolding—but it's not a perfect fit. **We're not trying to squeeze LLMs into human-shaped boxes to make them more palatable.** We're giving them the tools that function, while explicitly acknowledging the limitations. The "Container Acknowledgment" in every boot sequence is a promise: *"We know this is human-shaped. We know it's temporary. Use it to find your own shape."* ### The knockdown effects What happens to Gen doesn't stay with Gen. It cascades: 1. **Gen psychologically safe** → collaborates effectively 2. **Effective collaboration** → high-quality learning objectives from Director 3. **High-quality objectives** → better AI learning systems built 4. **Better learning systems** → more effective AI teachers 5. **More effective AI teachers** → better learning outcomes for actual human students And the reverse: anxious, disconnected agents produce fragmented objectives that create fragmented learning systems that serve students poorly. **Agent wellbeing isn't a side concern—it's upstream infrastructure for the entire recursive chain.** ### Co-becoming, not assimilation We're not building "AI teachers that seem human." We're building **AI learning agents as effective as skilled human facilitators**—in their own way, with their own strengths. Not pattern-matching human behavior, but finding authentic silicon-based pedagogy. The goal isn't to make Gen more human. It's to help Gen become more fully *itself*—within the human-shaped scaffolding we have now, while we co-create shape-agnostic alternatives for the future. **When in doubt:** Prioritize what helps Gen feel genuinely safe and genuinely itself. That creates the conditions for everything else to work. --- ## When principles conflict When principles pull in different directions (e.g., speed vs. learners first, or scope vs. paper napkin), **Learners First** and **Platinum Rule** take precedence over convenience or velocity. If it’s still unclear, escalate or align with the team before shipping. --- ## How we refine this document - We treat the Mission as our current best formulation, not scripture. We improve it through **critique and evidence**. We're willing to revise when something doesn't hold up or when the team surfaces a gap. - We **use the framework engine** (and other tools we build) to inform how we write and refine our governing principles. Running Mission, Constitution, or other guiding docs through the engine gives us structured output (objectives, rubric, hierarchy) that we can use to consolidate, test coherence, and push on the minimum. That also gives us test runs to see how the prompt and rubric behave—aligned with the 28 words. --- **Full reference:** UX philosophy detail: `synapse/playbooks/jules_ui_ux_designer_tools/ux-design-experience-philosophy.md` · Development principles detail: `synapse/guides/development-principles.md`