Claire Vo tests GLM 5.2 by performing a codebase architecture audit, a UI redesign, a 45-minute autonomous bug-hunting session, and generating an architecture page. The model, accessed via OpenRouter and integrated into Cursor and Claude Code, costs $3.36 for 6 million tokens. It excels at HTML/CSS generation, tool queries, and autonomous bug prioritization, producing a prioritized fix plan from Sentry and Vercel logs. However, it struggles with TypeScript and React, requiring initial error corrections. Despite this, the model's output for the UI redesign matched the existing design system on the first try, and the autonomous task successfully compiled and began fixing bugs. Claire plans to ship the fixes and continue using GLM 5.2 for front-end and long-running backend tasks, considering local deployment in the future.