Elgen's OS — Context-First Agent Workspace
Elgen's OS is how I work with AI day to day. Instead of starting every chat from zero, agents read the same written context (what I'm building, why, and the constraints) before they touch anything.
01. The Problem
Engineers lose massive context switching between terminals, linters, and conversational LLM interfaces. Generic AI assistants suffer from context amnesia, hallucinate non-existent APIs, and lack deterministic safety boundaries or long-term compounding memory.
02. Architecture & Solution
Every project gets a research folder with a numbered router doc that agents read first. A workspace agent (harnesses and scripts I run) keeps a memory bank of what I do across sessions, and a library of 1,000+ skills loads on demand instead of flooding the context window.
Key Architectural Highlights
- ▹Research folder + numbered router doc scaffolded into every project
- ▹Memory bank kept current by a workspace agent I run between sessions
- ▹1,000+ skills loaded only when a task needs them
03. Impact
Agents like Claude Code, Codex and Antigravity can work the same project in parallel without re-explaining it, and handoffs live in the repo instead of my head.
Elgen Mar Arinasa © 2026
Let's Discuss This Project →