Skip to main content
Normalized for Mintlify from knowledge-base/aiconnected-os/quick-system-overview.mdx.

​
What is aiConnectedOS?

Core Concept At its heart, aiConnectedOS is a virtual operating system for AI personas — not a generic agent platform. The fundamental differentiator is that it creates personalities with emotions, evolving consciousness-like qualities, and persistent identity that users “raise” rather than simply use. No two personas should ever be alike. Key Architectural Components The platform is built around several distinct layers. Instances serve as project workspaces containing chats, files, and assigned personas. Personas are the digital beings themselves — they learn, retain memories, develop mastery, and have fixed identities that evolve naturally over time. Cipher is the hidden orchestration layer that handles routing decisions, deciding which persona responds and when. CogniGraph is the knowledge graph powering persistent memory across sessions. Standout Features Multi-model routing is a defining capability — a single prompt can use different models for different jobs (one for research, one for writing, one for formatting). The platform also supports collaborative multi-persona chats, agentic teams (short-term, long-term, and recurring), and eventually a live browser window for research and agentic workflows. There’s also a Meeting Mode concept — recognizing that AI’s core problem is “not knowing when to shut up” in social contexts, so personas can act as passive ambient listeners in real-world situations. Business Model Pricing runs from free up to $99.99/month Pro tier, with higher tiers unlocking more instances, multi-model capability, persona creation, agentic teams, and the live browser. Bring-your-own-key is supported. There’s also a planned open source version as part of the 90-day launch plan. Design Philosophy The visual identity follows an OS-style aesthetic — dark backgrounds (#1e2328), Montserrat typography, thin icon rail with slide-over panels, and a mixed-mode light/dark interface. The benchmark is top-tier SaaS companies like Linear, Stripe, and Vercel. The design-to-code workflow embeds Adobe Illustrator SVGs directly in HTML with transparent interactive overlays on top. Build Roadmap An 18-week, 6-phase plan: Foundation → Chat Core → Instances & Files → Personas & Memory → Browser & Teams → Polish & Analytics. A LinkedIn announcement and launch campaign are also in preparation. It’s a genuinely ambitious platform — the combination of persistent identity, multi-model orchestration, and evolving personality is unlike anything currently in the market.

​
User Experience Overview

​
Getting Started

This is the first tab a new developer sees. It orients them before they touch anything technical.
  • Platform Overview — The zero-context explainer covering what aiConnectedOS is, the market problem it solves, what personas are, use cases, and the five-year vision. Written for investors and partners but equally useful for developer onboarding.
  • Vision & Philosophy — “Everything is conversational.” The north star, the “personalities not agents” thesis, and the reasoning behind every major architectural decision.
  • Brand & Identity — The aiConnected trademark position, the logo architecture (binary-encoded layers, hidden message), color palette, typography system, and naming conventions. References the Style Guide.
  • Glossary — Definitions for Cipher, Neurigraph, Instance, Persona, Skill Slot, Apprenticeship, Mod, and every other platform-specific term a developer will encounter.

​
Documentation Overview

​
Architecture

The structural layer. A developer reads this before touching any feature spec.
  • System Architecture Overview — How all layers relate: Cipher at the top, Neurigraph beneath, Personas as bounded expressions, Instances as workspaces, the System entity for OS-level commands.
  • Neurigraph Memory Architecture — The knowledge graph overview, the three memory layers (episodic, semantic, somatic), how nodes and embeddings work, and how memory flows through the system.
  • Multi-Model Routing Engine — How a single prompt can invoke different models for different subtasks, the Host Model concept, per-step model assignment, the Model Registry and Roles settings, and the orchestrator execution flow.
  • Fluid UI Architecture — The invisible OS paradigm, activity-based UI states, the event/ledger runtime, how Cipher stays hidden at the interface layer, and the ambient computing surface model across cars, glasses, mirrors, and robotics.
  • Fluid Context System — The chat-only context architecture: permanent context classes (topic, instructions, decisions, tone), the always-hot rolling window, and dynamic retrieval for older content.
  • Ambient Computing Vision — How personas maintain presence across all surfaces, the adaptation model per environment, and the long-horizon roadmap for embedded hardware.

​
Core Platform PRD

The numbered feature specifications, in build-order sequence. Each page maps to one numbered PRD part.
  • Instances & Dashboard — Instance as primary container, Instance types, the dashboard layout, the persistent open forum chat, the Instance-level settings layer, and the NEW action panel.
  • Personas System — Persona identities, skill slots, capability constraints, hard limits, persona-to-instance interaction, the Sally example (persistent memory across multiple instance assignments), and persona visibility controls.
  • Chat System — The chat kernel, thread types (forum/private/collaborative), message composition, streaming, system messages, tool output blocks, multi-persona participation, response routing, and the Chat Kernel as a reusable embed.
  • Search Architecture — Global search, Instance-scoped search, chat-level search injection, the unified no-copy-paste philosophy, search integration with the NEW action panel, and future persona-level and agentic search.
  • File & Document Management — Topic-level file systems, the general file system, per-topic inclusion toggles, per-file visibility (eye icon), separation of uploaded vs. generated content, file vectorization and graph linking, and global file search.
  • Live Documents — The per-instance document hub, the in-chat Live Document panel, AI-driven document manipulation commands, version history, export targets (Markdown, PDF, Google Docs), and the Document Chat mode.
  • Pin & Filter System — Message-level pinning, the filter bar (pinned/sent/received/media/links/search), the filtered export pipeline, and the Workspace/Instance board for promoted content.
  • Task System — Per-instance to-do list, message-to-task creation, task metadata, task-to-chat linking, email-from-task, Slack notify-from-task, the Task Agent (“what should I work on next”), and the on/off toggle per instance type.
  • Canvas & Whiteboard — Artifact bundles from chat selection, node types (message group, image, file, note, AI output), spatial arrangement, connection types, Canvas Chat mode, the link between pinning and canvas, and the three-view progression (list, board, graph).
  • Agentic Teams — The three team modes (short-term, long-term, recurring), the Executive Team structure (CEO/COO/CMO/CTO orchestrators), team creation and lifecycle, and integration with Instances and Personas.
  • Mods & Marketplace — Full-scale application modules, installation into the virtual Linux environment, persona access assignment, the mod store, third-party developer submission, the 20% revenue model, and approval controls.
  • Computer Use & Autonomous Browser — The autonomous development vision, extension vs. embedded vs. forked browser architecture options, reliability challenges, the path toward Manus-style autonomous execution, and integration with agentic teams.
  • Chat Cleanup & Bulk Operations — General chat as a staging area, mid-conversation move prompts, auto-rename suggestions, bulk multi-select across instances/personas/folders, Recently Deleted with 30-day retention, bulk memory operations.
  • Analytics & Insights — The Insights dashboard, persona utilization stats, hallucination refusal rate, reroute rate, time-to-resolution, user correction frequency, and the Phase 6 rollout plan.

​
Feature Specifications

Standalone features that extend the core platform. Each is a self-contained spec ready for engineering handoff.
  • Meeting Mode — One-tap persona behavioral state toggle, passive recording and transcription, wake-word-only activation during meetings, contextual memory integration, privacy considerations, and the “AI’s core problem is not knowing when to shut up” framing.
  • ChatNav: In-Chat Navigation — The sidebar table of contents, semantic and forced checkpoints, date-grouped sessions, hover/expand summaries, how checkpoint summaries serve as semantic routing metadata for the AI, and multi-persona onboarding via checkpoint walking.
  • ChatNav: Topic-Scoped Export (Addendum to ChatNav) — The ⋯ action menu on sidebar entries, the full option set (copy, download, move, share), the comparison table vs. Filter Bar Export, acceptance criteria, and implementation notes. Open decision: action menu threshold logic.
  • Dynamic Persona Waking — Wake-word activation per persona, the Last Active Device routing model, device hierarchy fallbacks, multi-persona simultaneous waking, graceful transitions, Meeting Mode integration, and the social etiquette layer for closing persona conversations.
  • Dynamic Screen Routing — Context-aware visual routing driven by Last Active Device signal, suppression logic for car and Meeting Mode contexts, verbal description as fallback, and deferred routing with reminders.
  • Conversation Split & Route — Drift detection, pre-selected inflection point with manual override, the non-destructive overlay notification (never part of the conversation record), Move vs. Copy decision, and memory attribution routing. Open decision: Move marker vs. clean removal.
  • Import & Migration — Inbound import from ChatGPT, Claude.ai, and aiConnectedOS native format; opt-in, non-destructive handling; read-only Import Archive instances; user-gated memory ingestion into Neurigraph; and Phase 6 placement.
  • Forget This / Memory Deprioritization — Graceful memory decay vs. hard deletion, the trust impact argument, the “This Was Just Thinking Out Loud” naming variant, and integration with Neurigraph’s deprioritization layer.
  • Automatic Conversation Cleanup — Live classification mid-conversation, the move prompt (yes/no), auto-rename suggestion flow, and the principle of general chat as an inbox.

​
Neurigraph

Deep-dive documentation for the memory architecture. Doubles as licensing-facing documentation.
  • Architecture Overview — The knowledge graph foundation, node types, embedding strategy, and how Neurigraph differs from simple vector stores or chat summaries.
  • Episodic Memory Layer — How individual experiences are stored, recalled, and aged over time.
  • Semantic Memory Layer — Concepts, facts, preferences, and domain knowledge as graph nodes.
  • Somatic Memory Layer — Motor pattern and embodied context memory for robotics and physical interaction.
  • Apprenticeship Model — Apprenticeships as standalone, open-ended, relational constructs (not pathways to skill slots), the distinction from skill acquisition, and the indefinite-by-design architecture.
  • Sleep & Dream Cycle — The mandatory 24-hour consolidation cron, what happens during sleep (compression, error correction, identity stabilization), the user-configurable sleep window, the fill-in system persona during unavailability, and why sleep cannot be interrupted.
  • Three-Tier Experience Stream — Unique experiences, common experiences (clustering threshold, cross-persona learning), guideline experiences (the Guidelines Layer as immutable safety instincts), the dream-time ingestion flow, and user opt-out controls.
  • Neurigraph for External Licensing — The partner-facing explainer: the AI memory continuity problem, licensing opportunities across gaming, medical, enterprise, and research, and ethical use constraints. This is the document written for prospective licensing partners.
  • Patentability Assessment — Summary of the strongest patent claims around the integrated episodic/somatic/semantic/compounding memory architecture for embodied systems, and the recommendation for patent counsel.

​
Personas

A focused section on the Personas system, separated from the core PRD for depth.
  • Personas Overview — What personas are, why they are “personalities not agents,” the Tamagotchi analogy for individual investment, and the Virtual Villagers analogy for multi-persona dynamics.
  • Skill Slots & Capability Constraints — Fixed slot count, slot categories, capability enforcement, inline refusal with explanation, Cipher escalation for ambiguous cases, and the “what this persona can help with” UI transparency layer.
  • Persona Lifecycle — Creation, training, sleep, growth, and the rules of personhood (no overwritten memories, no identity drift, layered growth only).
  • Virtual Employee Model — Micro-experience accumulation, working hours and response latency realism, voice and email presence, the employer-employee framing, and how Sally remembers Frank’s client project six months later.
  • Pixel Agents, Tamagotchi & Virtual Villagers — The cultural reference analysis: why Pixel Agents-style task visualization is the wrong layer, why Tamagotchi is the most accurate emotional investment analogy, and when to use Virtual Villagers for non-technical communication.

​
Robotics

The “CarPlay for Robotics” layer. Partially documented; some sections still need formal writeup.
  • Robotics Strategy Overview — The three-layer stack (aiConnectedOS as universal cognitive layer, manufacturer SDK, developer extensions), the CarPlay analogy, and the strategic position.
  • Certification Tiers — Platform-defined Level 0-3 + Level X, why tiers must be platform-defined not manufacturer-defined, and the initial default capability suite (~30 capabilities).
  • Robot Class Taxonomy — Humanoid, industrial/manufacturing, aerial, mobile platform, companion/stationary, and why class definitions are required for certification precision.
  • Governance, Liability & Revenue (needs formal writeup) — How analogous industries handle liability (auto, app stores, medical device Notified Bodies model, FAA drone regulations), the open Level 2 liability scenario, certification revenue model options, and international regulatory fragmentation.

​
Build Roadmap

The engineering execution plan.
  • 18-Week Build Plan Overview — The six-phase structure, phase gates, and the principle of shipping the runtime before the adapters.
  • Phase 1: Foundation (Weeks 1-3) — Project scaffolding, auth, basic UI shell.
  • Phase 2: Chat Core (Weeks 4-6) — The complete messaging system, streaming, AI responses.
  • Phase 3: Instances & Files (Weeks 7-9) — Workspace management and file handling.
  • Phase 4: Personas & Memory (Weeks 10-12) — The persona system and Neurigraph integration.
  • Phase 5: Browser & Teams (Weeks 13-16) — Collaborative browsing and agentic team orchestration.
  • Phase 6: Polish & Analytics (Weeks 17-18) — Insights dashboard, Import & Migration, production deployment.

​
Design System

Reference material for any developer touching the UI.
  • Style Guide — Color palette (#1e2328 background, #839aac text, #2e95f3 accent, full navy scale), typography (Montserrat primary, DM Sans secondary), spacing, and component conventions.
  • Welcome Screen — The Illustrator-sourced reference design with annotated layout regions, the OS-style menu bar, navigation rail, stat panel, recent conversations, and persona dock.
  • Design-to-Code Pipeline — The mandatory rule: embed Illustrator SVG directly in HTML, layer transparent interactive overlays on top, never rebuild or reinterpret from extracted specs. The rationale for keeping design iteration in the chat interface and engineering integration in Claude Code.
  • React Prototype — The working aiConnectedOS-prototype.jsx file with component map, known icon substitutions for the artifact environment, and screen inventory.

​
Open Items

A living page. Updated before each engineering handoff.
  • Decisions Pending — ChatNav action menu threshold logic (⋯ on all entries vs. entries above N messages). Conversation Split & Route: Move marker vs. clean removal.
  • Specs Needing Formal Writeup — Robotics governance and liability framework. Persona sleep cycle as standalone PRD section. “Forget This” as standalone PRD section. Brand identity and trademark strategy document. Virtual Employee presence layer (voice, email, working hours).
That is the complete structure. Every document you have goes into one of those slots, and every open item has a named home waiting for it.
Last modified on April 18, 2026