Personalized Multi-Agent AI Assistant: Pacific Slate
A self-updating, data-preserving AI assistant that simplifies daily tasks.
1The Birth of Pacific Slate
An MBA student, overwhelmed by internship applications, family responsibilities, and information overload, grew tired of constantly re-explaining his situation to changing AI tools. This need led to the design of a self-updating system that filters noise and keeps data secure. The system runs on a rented server, holding encrypted backups of important files.
2Design Philosophy
The system is built on durability, utility, and productivity. It's not a product or research project but a daily-use tool. Designed to function without user intervention, initial mobile and desktop clients were abandoned as unsustainable. Instead, it integrates with common AI development tools.
3Technical Infrastructure
Eight specialized agents work on models optimized for their tasks: one main agent for routing requests and seven experts for research, coding, analysis, etc. Each agent runs on a task-specific model, optimizing resource use. Responses include sources and costs for transparency.
4Data Control and Security
The user retains full control over accumulated data, with nothing becoming vendor property. Data is exportable and deletable, never used for training. Zero-data-retention endpoints are employed, and work can be redirected to server-hosted models when needed.
5Open Source and Flexibility
Most components are open-source, running on a personal server. Open-source alternatives are preferred when available, with only model rental and one swappable memory service requiring payment. The design prioritizes adaptability.
6Development Journey
Without a software engineering background, the creator leveraged online tools to build this system. The architecture was independently designed: components, connections, failure modes, and budget management. AI agents wrote most of the code through specification and review processes.
7Operational Insights
In operation since early 2026, daily use revealed most technical details. For instance, the backup layer was added after discovering the framework's native backup settings were ineffective. Most issues were design gaps rather than bugs, uncovered only through real-world operation.
8Future-Proof Design
Model-agnostic architecture allows easy adoption of better models. The system works autonomously: scheduled tasks, low-risk dependencies, self-merging security updates. Failures degrade to working states without waiting for user intervention, ensuring continuous utility.
Original article: Hacker News: Show HN — Kişiselleştirilmiş Çoklu Yapay Zeka Asistanı: Pacific Slate