I Built a Local AI That Knows Who I Am and Runs in My Basement

2026-04-26 · Source: Machine Learning on Medium · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Cybersecurity & Data Privacy, Software Development & Engineering · Depth: Intermediate, quick

Summary

A developer has created "Oracle," a local, offline AI system running on a Raspberry Pi CM5 in their basement. Oracle is built on "Native Compression Intelligence" (NCI), an architecture that stores knowledge as interconnected concepts rather than neural network parameters, enabling it to remember personal context like the user's name, family, and projects. The system currently has over 10,000 trained knowledge nodes and operates without cloud dependencies or subscriptions. A key feature is its five-layer Ethics Engine, with a cryptographically sealed child safety layer that cannot be modified without preventing the system from starting. While not yet as powerful as commercial AI services like Claude or ChatGPT, Oracle represents a personalized, privacy-focused alternative, utilizing Gemma 3 12B via Ollama for language generation and a custom programming language called Terse for its rules.

Key takeaway

For AI Engineers and hobbyists exploring personalized, privacy-centric AI, you should consider developing local systems that integrate custom knowledge architectures with open-source models. Focusing on hardware ownership and embedded ethical frameworks, like Oracle's cryptographically sealed safety layer, can mitigate privacy concerns and ensure system integrity, offering a path toward truly autonomous and user-controlled AI.

Key insights

Personalized, local AI systems can offer enhanced privacy and customizability beyond commercial cloud services.

Principles

• Knowledge as interconnected concepts.
• Cryptographic integrity for ethical safeguards.

Method

The Oracle system integrates a custom NCI brain with a Gemma 3 12B language model via Ollama, running on a Raspberry Pi CM5 with OPNsense for local network operations, and governed by a five-layer Ethics Engine.

In practice

• Run Gemma 3 12B locally via Ollama.
• Use Raspberry Pi CM5 for edge AI.
• Implement OPNsense for local network.

Topics

• Local AI
• Native Compression Intelligence
• Ethics Engine
• Raspberry Pi CM5
• Offline AI

Best for: AI Engineer, Machine Learning Engineer, AI Architect

Related on AIssential

• I Taught My Local AI to Read Everything I’ve Ever Written — Local AI can privately learn from personal data using simple, self-hosted automation.
• Perplexity’s Personal Computer is now available to everyone on Mac — Perplexity's Personal Computer brings secure, local AI agent capabilities to Mac users for multi-step workflows.
• Stanford Researchers Release OpenJarvis: A Local-First Framework for Building On-Device Personal AI Agents with Tools, Memory, and Learning — OpenJarvis provides a local-first, modular framework for building on-device personal AI agents with integrated tools and memory.
• How To Use AI On Your Phone WITHOUT Internet — Locally AI enables private, offline AI model execution directly on user devices, ensuring data privacy.
• 🔮 Why I changed my mind about Apple — Apple's hardware, not its cloud AI, is becoming the preferred platform for local AI inference due to unique architectural advantages.
• Local AI — Local AI models are becoming viable alternatives to cloud APIs due to cost, privacy, performance, and fine-tuning capabilities.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Machine Learning on Medium.