PERSON DIRECTORY
John Yang
John Yang appears in 2 indexed conversations across Latent Space, 十字路口Crossing. This directory brings every appearance, source, TL;DR, digest, and transcript into one searchable feed.
人类和 AI Agent 的最佳配合方式，还没被发明｜对谈 Paperboy
Paperboy押注在OS层观察用户，通过屏幕、键盘和跨应用活动构建可持续的agent记忆。团队以持久可搜索的聊天流替代session和prompt，试图提升个性化、主动服务与跨应用协作能力。屏幕数据采集正成为行业共识，但pipeline尚无最佳方案；Paperboy目前融资$4.7M、每天亏损，下一轮产品迭代与成本改善值得跟踪。
[State of Code Evals] After SWE-bench, Code Clash & SOTA Coding Benchmarks recap — John Yang
SWE-bench became a commercial catalyst after Devin’s “mind-blowing” launch, while CodeClash shifts evaluation toward repeated development of consequential codebases rather than one-shot patches.The emerging benchmark portfolio spans cost, domains, and autonomy, but impossible tasks, human–AI interaction data, and realistic simulators remain critical tests of whether leaderboard gains translate into durable workflows.