Four Drawers, One Agent
CLAUDE.md is not memory, a vector search is not a decision, and OpenViking is not a nicer RAG. Here is which drawer holds what.
Tech Talk with Veeresh
Principal QA Architect · AI Test Architect · writing on AI-driven test strategy, automation frameworks, and quality engineering teams. About Veeresh lives on the portfolio.
Latest Posts
All 40 postsWeb-Grounded Answers Without a Server: What Chrome's On-Device AI Can (and Can't) Do
A portfolio chatbot needed to answer questions outside its own knowledge base without shipping a search API key to every visitor. Scoping that feature surfaces a real architectural lesson about...
Self-Hosting an AI Chatbot on a $150 Raspberry Pi 5
No cloud bill, no vendor lock-in, no rate limits, just Nginx, PM2, and a local LLM running 24/7 on a computer the size of a deck of cards. Here's the...
Self-Healing Test Suites: AI-Powered Locator Healing in CI/CD
Frontend renamed a button and your suite went red at 2am? Self-healing locators find it by what it does, not what it's called.
Playwright AI Codegen in 2026: Generating Test Suites from Natural Language
Playwright's codegen --ai writes tests from plain English. I still review everything — but I type about 80% less than I used to.