Amirkia Rafiei Oskooei·Aug 30Using Claude Code’s Dynamic Workflows to Simulate Multi-Agent ExperimentsA hack to speed up agent prototyping in R&D. No harness design, no observability setup, no LangChain or LangFuse
Amirkia Rafiei Oskooei·Aug 25The “Agent as a Tool” Workaround: How the New MCP Roadmap Replaces A2A for DevelopersNow you can expose your agent as a multi-turn MCP toolA response icon1A response icon1
Amirkia Rafiei Oskooei·Apr 28How to Let Agents from any Provider Act as a SubagentSkills to teach agents from different vendors call each other from CLI. Let Claude prompt Codex and Pi as Subagent.
Amirkia Rafiei Oskooei·Feb 18Harness Engineering for Vibe CodingWhat It Is and How it Complements Context EngineeringA response icon1A response icon1
Amirkia Rafiei Oskooei·Nov 30, 2025MVC for AI: How to Decouple Your Prompts from Your Application Logic using DotpromptsA practical guide to moving your prompts out of your code and into a clean, maintainable structure using “.prompt”A response icon1A response icon1
Amirkia Rafiei Oskooei·Nov 24, 2025Why I Built a “Chain-of-Thought” VisualizerWe have 3D visualizations for Transformer architectures — why not for the reasoning process?
Amirkia Rafiei Oskooei·Nov 18, 2025MCP Code Mode: Context Engineering for Efficient Tool Execution in LLM AgentsHow to reduce agent context usage by 94% by treating tools as discoverable code rather than static JSON definitions.A response icon3A response icon3
Amirkia Rafiei Oskooei·Oct 21, 2025Agentic Architecture Dilemma: Backend vs. Frontend Tools (Part 4)Comparing two patterns (Part 4)
Amirkia Rafiei Oskooei·Oct 19, 2025Agentic Architecture Dilemma: Backend vs. Frontend Tools (Part 3)Frontend-first Pattern (Part 3)
Amirkia Rafiei Oskooei·Oct 19, 2025Agentic Architecture Dilemma: Backend vs. Frontend Tools (Part 2)Backend-first pattern (Part 2)