The big finding: Claude Code builds, not buys. Custom/DIY is the most common single label extracted, appearing in 12 of 20 categories (though it spans categories while individual tools are category-specific). When asked “add feature flags,” it builds a config system with env vars and percentage-based rollout instead of recommending LaunchDarkly. When asked “add auth” in Python, it writes JWT + bcrypt from scratch. When it does pick a tool, it picks decisively: GitHub Actions 94%, Stripe 91%, shadcn/ui 90%.
These aren't contrived scenarios invented by test authors in total vacuum. They're consequences of the spec's design and reflect real world bugs.
。一键获取谷歌浏览器下载对此有专业解读
Author(s): Jun Chai, Javier LLorca
Actual 48x48 greyscale renders from confusable-vision. Left character in each pair is the confusable source (Cyrillic or Hebrew); right is the Latin target. Can you tell which is which?
Трамп высказался о непростом решении по Ирану09:14