Hyunjoong Kim's Technology Blog
Practical writing on AI agents, SDLC documentation automation, full-stack architecture, and product validation.
31 results
An Operating Model for AI Agent Sandboxes: Permissions, Networks, and Worktrees
To use coding agents more often, teams do not need heavier supervision. They need clearer boundaries. This guide turns permissions, network access, worktrees, and CI gates into a practical sandbox operating model.
Delegation Packets for AI-Native Teams: The Small Spec Coding Agents Need
Claude Code, Codex, and other coding agents do not need longer prompts as much as they need better delegation packets. This practical guide shows how to define outcome, scope, permissions, verification, and stop conditions before handing work to an agent.
How to Design Agent Runbooks for AI-Native Development Teams
A practical guide to goals, permissions, verification, and escalation rules for teams adopting coding agents such as Claude Code and Codex.
The real bottleneck in the MCP era is not the model, it's permissions
As AI agents become better at using tools, the bottleneck shifts from model quality to permission design. This post lays out a least-privilege operating model for MCP-based tool use.
What Is a Vector Database? Why It Matters for AI Search and RAG
A practical explanation of vector databases, how they differ from traditional databases, and why they matter for AI search, RAG, and recommendation systems.
How Vector Databases Work: Embeddings, Retrieval, and Matching in Practice
A technical explanation of how vector databases store embeddings, search them with ANN, and match results back to original documents.