https://tengli.dev Teng Li — ML Engineer https://tengli.dev/posts/calibrating-mcp-evals.html https://tengli.dev/posts/calibrating-mcp-evals.html Mon, 27 Jul 2026 00:00:00 GMT Three rounds of calibration, four real servers, $0.60 in API costs — how I made an LLM-powered tool-selection benchmark fair enough to publish, and what it revealed about static lint rules predicting live model behavior. https://tengli.dev/posts/mcp-servers-failing-agents.html https://tengli.dev/posts/mcp-servers-failing-agents.html Tue, 21 Jul 2026 00:00:00 GMT Your MCP server can be 100% spec-compliant and still be unusable by an agent. I built mcpgrade to measure the difference — then pointed it at 36 popular servers. https://tengli.dev/posts/tool-design-for-llms.html https://tengli.dev/posts/tool-design-for-llms.html Sun, 12 Jul 2026 00:00:00 GMT Your tool API can be perfectly engineered and still confuse every model that touches it. Field notes on why, from someone who integrates agent tools for a living. https://tengli.dev/posts/hello-world.html https://tengli.dev/posts/hello-world.html Sun, 12 Jul 2026 00:00:00 GMT Production AI agents fail in ways that never show up in demos. Someone should write those down.