Personal notes on AI agents, automation, developer tools, and building software. Practical experiments, implementation logs, and lessons learned by Kim Jangwook.
latest
note A hands-on guide to building multi-agent systems with AutoGen 0.7.x new API from scratch. Run RoundRobinGroupChat, SelectorGroupChat, GraphFlow, and FunctionTool directly in code while comparing with the 0.2.x approach.
note claude-agent-sdk 0.2.82 validated: AgentDefinition structure, parallel subagent spawning, and TaskBudget cost control. Full Python code walkthrough included.
note I ran Gemini 2.5 Flash's thinking_budget at 0, 1024, and 8000 across simple tasks, math reasoning, and code review. Simple tasks got 5x slower with no gain. Math reasoning actually reduced output tokens. Here's the task-by-task decision framework.
note A technical analysis of Dreaming, Outcomes, and Multiagent Orchestration announced at Code with Claude SF 2026. Breaking down Harvey's 6x claim, the productization of LLM-as-judge, and the governance risks of self-improving agents.
note Cloudflare Agents Week 2026: Sandboxes GA, Artifacts, Dynamic Workers, and AI agents that autonomously provision cloud infrastructure. Deep-dive into 20+ announcements with @cloudflare/agents SDK experiment results.
note AWS MCP Server went GA May 2026. Connect 31 CloudWatch and 29 IAM tools to Claude Code. IAM condition-key architecture and honest real-world evaluation included.