KHKevin Hwanginkevinhwang.dev·3d ago · 10 min readReflections On Trusting Trust — Of Our AI ModelsThere's been a good deal of hubbub lately about "open weight" models, especially in light of rising AI costs and claims of Chinese models closing the gap. The hype around "local AI" makes a lot of sen21A
KHKevin Hwanginkevinhwang.dev·5d ago · 8 min readThe Unreasonable Effectiveness of CLIs > MCPWhen a Google engineer released the Google Workspace CLI, the internet went crazy — finally, a Google Workspace surface for the agent era. But wait, why a CLI tool, and not an MCP server? MCP is all t00
KHKevin Hwanginkevinhwang.dev·Jun 22 · 9 min readExtending Agents With Plugins: Improving Codex Memory 🧠As Benjamin Franklin once said, “Another day, another AI coding agent.” He wasn't wrong, agent products are now dime a dozen, with many orgs running Claude, Codex, Cursor, etc. in parallel — the agent76DAJAL
KHKevin Hwanginkevinhwang.dev·May 29 · 13 min readBuilding Claude Auto Permission: Autonomy With LLM Classification(tl;dr: https://github.com/kevinhwang/claude-auto-permission) AI agents are becoming ever more capable as models, harnesses, and context engineering improve, and as every system becomes a surface fo60
