Hey HN. Coding agents write code fast. Nobody verifies it works. We use Cursor and Claude Code daily. The pattern is always the same: code passes CI, types check, linting clean – but the app breaks in the browser. Auth flows fail silently. Modals don't close. MetaMask popups don't connect. CSS overlaps in production. CI was built for human-written code at human speed. Agentic development generates 10x more code, 10x more PRs, and CI catches maybe 20% of the real issues. The rest – visual regressions, broken user flows, state bugs, wallet interactions – only show up when someone actually uses the app. Agentiqa is the verification layer for this era. An AI QA agent that tests your app the way a user would: in a real browser, on mobile, through Web3 wallet flows. No test scripts. Give it a URL, it explores autonomously using vision models, generates a test plan from what it sees, executes in Playwright, and reports what's broken with screenshots and repro steps. What we cover: web apps, mobile (responsive + native), and DeFi/Web3 (MetaMask, WalletConnect, tx signing – nobody else does this end-to-end). When the UI changes, the agent adapts because it understands intent, not selectors. 2 founders, 6 months, 185 users, free tier. Built for teams shipping with coding agents. What are we missing?