# Zomni Solutions

> Zomni is an AI product and research company pioneering persistent agentic intelligence and transformational companions designed for real human outcomes and cognitive sovereignty. Recipient of startup grants from OpenAI and Google Cloud.

## Overview

Zomni builds human-centered AI systems that perceive, decide, and act with precision, purpose, and restraint. Rather than engineering addictive parasocial loops or screen-stickiness, Zomni operates on a strict foundational mandate: across time and repeated interactions, the human must become more capable and self-reliant—never dependent.

- **Foundational Thesis**: "Life moved, not screen time." The true measure of intelligence is verified physical-world action, preserved human authorship, and decision clarity.
- **Flagship Consumer Product**: [Cosmuna](https://cosmuna.com) — A personal transformation companion engineered with persistent memory, natural timing, and verified physical proof ledgers.
- **Products & Frameworks**: Fine-tuned intelligence, state-driven agentic frameworks, persistent memory architectures, and applied executive systems.
- **Research Moat**: The Living State Machine (LSM) dynamical state engine, the Human Agency Evaluation Suite (HAES), and formal agentic restraint policies.

## Core Products & Platforms

- [Products & Frameworks](https://zomni.ai/products): Zomni's complete portfolio of fine-tuned intelligence, state-driven agentic frameworks, and persistent memory architectures. Showcases Cosmuna (consumer companion), the 3-tiered Living Memory Stack, Contextual Restraint & Timing Engines, Intent-Preserving Agent Mesh, and Applied Solutions.
- [Cosmuna](https://zomni.ai/cosmuna): The flagship transformation companion. Features The Blueprint™ ($B(t)$) engine, 3-tiered memory (PostgreSQL + pgvector + immutable Proof Ledger), C-bound Swiss Ephemeris astronomical engine, and specialized Mastra agents.
- [Research](https://zomni.ai/approach): The foundational scientific paradigm, Living State Machine formulation ($B(t) = T(t) \otimes H(t)$), research programs, and three-horizon grant roadmap. Accessible at `https://zomni.ai/research` or `https://zomni.ai/approach`.
- [About Zomni](https://zomni.ai/about): Company origins, leadership (Raj K, SKS), core values, institutional support (OpenAI, Google Cloud), and the dual-engine company model.
- [Research & Working Papers](https://zomni.ai/resources): Working papers covering Human Agency Evaluation, Temporal Context Graphs, Agentic Restraint, and Outcome Attribution. Features the formal 2-page Executive Concept Paper & Grant Expression of Interest (EOI) for the Human Agency Evaluation Suite (HAES) (6-Month Open-Source Research Initiative).
- [Headquarters](https://zomni.ai/hq): Spatial console, live concierge desk with Zus (Chief Happiness Officer), engineering telemetry, and agentic squad operations.
- [Contact & Partnerships](https://zomni.ai/contact): Institutional partnerships, research collaborations, investor inquiries, and founder discovery meetings (Direct calendar: `https://cal.id/zomni`).

## Key Terminology & Proprietary Primitives

- **The Living State Machine (LSM)**: The overarching architectural paradigm of continuous, dynamical state-space AI that synthesizes continuous temporal drivers with evolving human state models to preserve human agency without memory drift.
- **The Blueprint™ Engine ($B(t)$)**: Zomni’s proprietary, trademarked state-vector primitive that implements the Living State Machine in software ($B(t) = T(t) \otimes H(t)$), holding the living mathematical synthesis of temporal dynamics and evolving human identity.
- **Life Moved, Not Screen Time**: The non-negotiable north star metric measuring verified real-world follow-through, unblocked friction, and honored commitments over time-in-app.
- **3-Tiered Memory Architecture**: An anti-hallucination, anti-drift memory stack combining Zod-validated structured working memory in PostgreSQL, semantic pgvector recall, and an append-only Proof Ledger.
- **The 5 Restraint Primitives**: An agentic intervention policy dictating when an AI must *Ask*, *Suggest*, *Execute*, *Defer*, or *Remain Silent* to protect human solitude and self-determination.
- **Human Agency Evaluation Suite (HAES)**: Empirical testing suite and open-source benchmark (repository: `https://github.com/apprider/haes`) that measures whether persistent AI systems strengthen user critical judgment or induce learned helplessness via the Agency Preservation Score ($APS$).
- **Temporal Context Graph (TCG)**: Dynamic memory architecture separating past behavioral patterns from what the user is actively striving to become, preventing identity lock-in.


## External Profiles & Links

- **Flagship Application**: [Cosmuna Website](https://cosmuna.com)
- **Investor Profile**: [Peachscore Profile](https://app.peachscore.com/company/zomni-solutions-private-ltd)
- **Primary Domain**: [https://zomni.ai](https://zomni.ai)
- **Contact**: [contact@zomni.ai](mailto:contact@zomni.ai)
- **Headquarters**: London, United Kingdom

---

For the comprehensive, full-context architectural documentation, research papers, and technical specifications, see [llms-full.txt](https://zomni.ai/llms-full.txt).