Notes on AI QA: practical testing guides, engineering decisions behind Prufa, and QA for agent-driven development.https://prufa.dev/https://prufa.dev/blog/guides/qa-for-vibe-coded-apps/https://prufa.dev/blog/guides/qa-for-vibe-coded-apps/What actually breaks in vibe-coded web apps, in what order, and what to do before launch. Data from 49 fresh launches, June 2026.Fri, 12 Jun 2026 00:00:00 GMTguideshttps://prufa.dev/blog/agents/llms-txt-who-actually-reads-it/https://prufa.dev/blog/agents/llms-txt-who-actually-reads-it/Major AI crawlers don't fetch llms.txt — coding agents do. The honest state of the spec in 2026, and the 30-minute version that's still worth shipping.Thu, 11 Jun 2026 00:00:00 GMTagentshttps://prufa.dev/blog/engineering/we-audited-49-show-hn-launches/https://prufa.dev/blog/engineering/we-audited-49-show-hn-launches/Free QA audits on 49 fresh Show HN launches: 78% had a critical finding. The full breakdown of what actually breaks at launch — analytics, links, cookies.Thu, 11 Jun 2026 00:00:00 GMTengineeringhttps://prufa.dev/blog/guides/website-qa-checklist-before-launch/https://prufa.dev/blog/guides/website-qa-checklist-before-launch/A pre-launch QA checklist ordered by real failure data from 49 audited launches: analytics first, then forms, links, SEO basics, consent, and user flows.Thu, 11 Jun 2026 00:00:00 GMTguideshttps://prufa.dev/blog/ai-qa/why-prufa-exists/https://prufa.dev/blog/ai-qa/why-prufa-exists/AI writes 10× more code than humans can validate. The QA gap isn't a missing test suite — it's surface area. Here's the architecture we bet on.Thu, 11 Jun 2026 00:00:00 GMTai-qahttps://prufa.dev/blog/engineering/how-prufa-verifies-a-signup-flow/https://prufa.dev/blog/engineering/how-prufa-verifies-a-signup-flow/One signup run, end to end: an LLM-backed agent drives the browser, plain code grades the outcome against a public flow-spec. Same input, same verdict.Thu, 11 Jun 2026 00:00:00 GMTengineering