Founders Wei-Wei and Jeff are launching Momentic (YC W24), QA agents that autonomously plan, explore, and bug-bash controls and user flows across web, iOS, Android, and soon desktop apps. The pair previously worked on dev tooling and infrastructure at Robinhood, Retool, WeWork, and Qualtrics, and built Momentic after finding Playwright and Selenium painful to maintain.
Their custom agent harness, Mo, uses a self-organizing agent graph plus specialized planner and locator agents. In a 60-minute benchmark across Gitea (58.8k stars), Metabase, Twenty (YC), and Saleor with 1,396 human-labeled bugs, Mo found 269 bugs with 19.3% recall and 84% precision at $2.67 per bug.
That outpaced Codex (86 bugs, 6.2% recall, $0.80 per bug), Claude Code (59 bugs, 4.2% recall, $4.09 per bug), and Devin (27 bugs, 1.9% recall, $3.89 per bug). Momentic argues the future of QA is agents testing themselves rather than writing scripts.
Source: Hacker News · Summarized by HeadlinesBriefing