Blog

Notes on building qualien.ai, as they happen. This is a new project, so there’s just the one post so far.

July 31, 2026

Why I’m building qualien.ai

I work in software, and like a lot of people right now, I’ve been using AI coding assistants daily. They’re genuinely good at writing code that looks like a test. What they’re not good at is understanding testing itself — coverage, flakiness, the actual conventions of frameworks like Playwright or Selenium, or when a “passing” test is quietly a bad one.

That gap is where qualien.ai started. Not as a company, not as a product with a launch date and a roadmap — just a personal project to see how far I could get building something that treats real testing knowledge as a first-class input, instead of an afterthought bolted onto a general-purpose coding assistant.

Grounded, not generic

The one idea I keep coming back to: instead of letting a model freestyle a test case or a framework setup from whatever it remembers, ground the answer in real, checkable reference material — retrieved, re-ranked, and cited. That’s why “grounded” shows up everywhere on this site. It’s not a marketing word here — it’s the actual mechanism, and it’s written up in more detail on the documentation page.

Built one piece at a time

qualien.ai grew in phases: a chat assistant grounded in testing docs first, then single-purpose generators for test cases, framework scaffolds, automation code, and code review. Then Olympus — a multi-agent system that runs a full requirements-to-report test cycle against a real, live application, with a real browser behind it. Most recently, Relay — the same plan/generate/heal loop, but running inside your own AI assistant over MCP, so it doesn’t use a single qualien credit. Next is Pulse — extending that same grounded, agentic approach to native mobile apps on real iOS and Android devices.

Where this is headed

I’m not pretending this is a finished platform — it’s a young, evolving personal project, and I’d rather say that plainly than oversell it. If you write or maintain test suites and want AI help that actually understands testing, rather than just producing code that resembles it, that’s exactly who I’m building this for.