Engineering notes

Behind-the-scenes on building autonomous AI agents with MCP.

June 4, 2026 · gashel01

I Benchmarked My Own Libraries Against LlamaIndex, Mem0, Tavily, DeepEval and a Raw Sandbox — Honestly

Five standalone libraries, five named incumbents, the same rule: control the confounds, run the competitor at its best, publish the losses as plainly as the wins. Spoiler — none was a clean sweep, and every run found a real bug.

June 3, 2026 · gashel01

I Benchmarked My Agent Framework Against LangGraph and CrewAI — Honestly

Same model, same tools over MCP, reproducible runs. Here's exactly where my kernel wins, where it ties, and where it loses — including the toy task where the bare agent beats it.

April 20, 2025 · gashel01

Stop Letting Your LLM Drive: Compile the Plan, Execute It Deterministically

LTP treats the LLM as a compiler, not a runtime — one call to plan, then deterministic Python to run it. Here's how it works, and the honest, reproducible numbers (it's not a silver bullet).

April 15, 2025 · gashel01

I Built a 10-Library AI Agent Suite on MCP — Here's What I Learned

Ten Python libraries, 120+ tools, one orchestration kernel — all speaking MCP. This is what I learned building the MCP AI Suite.