Projects and Writings
Projects
A RAG-powered Q&A tool built on the AI That Works podcast archive. Ask questions, get answers with source citations from the show.
Writing
May 20, 2026
TL;DR: Single-method retrieval has limited levers to enhance your application. A hybrid retrieval pipeline with cross-encoder gives you the ability to tune precision and recall by casting a wide net first and narrowing aggressively. This ar...
May 7, 2026
Generated with Gemini TL;DR How I built an MVP knowledge navigator using a Spec-Driven Development approach. The build took roughly 1–1.5 weeks from initial research to local verification, plus another week to handle hosting (i.e. figuring ...
March 9, 2026
TL;DR: How you use AI can help or hinder how you learn new skills. A recent Anthropic study shows a 62 point mastery gap predicated on how you use AI. Users who stayed cognitively engaged learned new skills; those who delegated didn’t. This...
February 5, 2026
TL;DR: Proactive evals catch issues before users do Most teams build evals reactively — wait for failures, scramble to add tests, repeat. This article shows a proactive way: map user journeys, generate test cases with LLM assistance, valida...
January 2, 2026
The bottom line: The faster AI writes code, the more your job becomes defining what to build. I’ve been using Antigravity — an AI asssited coding tool — since launch day. Before that, I used Windsurf. Both let AI agents to write code autono...
November 28, 2025
Generated with Gemini (Nano Banana) Executive Summary Most organizations build AI evaluations reactively — testing only for failures they’ve already experienced. This survivorship bias leaves systems vulnerable to unforeseen problems, espec...