Writing · 2026

Engineering notes from the inside.

Production-grade deep-dives on AI/LLM systems I've shipped — what I built, what I learned, and the calls I'd defend in code review.

May 12, 2026 · 9 min read

Building a Zero-Dependency PDF Parser to Survive Vercel Serverless

How I dodged DOMMatrix is not defined and shipped a RAG agent that actually deploys.

RAGAnthropic ClaudeNext.jsServerless
May 17, 2026 · 18 min read

Building Talk to My Portfolio: An MCP Server + Embeddable AI Widget

How I shipped a chat widget that lets visitors interrogate my CV, projects, and 164 Platzi certificates — and the six production bugs I had to fix to get it working on Vercel.

MCPAI AgentsAnthropic ClaudeNext.jsPreact
May 19, 2026 · 11 min read

Three AI Agents in Parallel + One Supervisor

How Promise.all, Zod, and Server-Sent Events compose into a multi-agent code reviewer.

Multi-agentAnthropic ClaudeVercel AI SDKZod
May 26, 2026 · 10 min read

The Cleanest Pattern I've Found for Autonomous Agents: A Terminal Tool With No execute

Why ToolLoopAgent + a no-execute finalAnswer tool replaced 80 lines of brittle while-loop code in my research agent.

AI AgentsAnthropic ClaudeVercel AI SDKToolLoopAgent