Articles & Insights
Deep dives into the QA journey, from building production apps with dual AI fleets to evolving a monolithic skill into a multi-agent system.
I Built a Production PWA in 5 Days. One AI Fleet Built It. Another's Job Was to Break It.
Here's what happened when I put two AI agent systems in the same codebase with opposite goals.
SafePetPlate is a free app that lets pet owners instantly check if a food is safe for their dog or cat. I built the entire thing in 5 days using two AI agent fleets: Claude Flow (the builders) and Agentic QE Fleet (the quality enforcers).
From One Giant Skill to an AI-Powered QA Team
How We Built an Agentic Quality Engineering System
A story about learning, iterating, and integrating AI agents into a real QA team. From one oversized skill to a fleet of specialized agents, guardrails, and orchestrator pipelines, combining human judgment with AI efficiency.
How I Onboarded Myself Into a Production App Using Exploratory Testing and Playwright MCP
Solo QA Audit: Exploratory Testing & Browser Automation
No documentation. No handover. Just a URL. Here's how I used exploratory testing, Playwright MCP, and Notion to audit an AI chatbot I'd never seen before, and found 37 bugs.
Taming Playwright MCP: Five Ways to Cut Token Costs
Keeping the Live Browser, Dropping the Token Bill
A single locator hunt was costing more tokens than the rest of my QA pipeline combined. Five fixes changed that: a locked-down toolset, targeted DOM queries that save around 90%, hard-stripped fallback snapshots, and moving debug work off the browser entirely, without giving up the live browser.
Most Fitness Apps Replace the Trainer. We Kept Him.
Product and QA on a Real Client Build
The exercise library in this app is admin only and no member will ever see it. How a website brief for a small gym in Novi Sad became an online training business, what we refused to build, and the three bugs that passed every automated check.