四个真实公司给AI Agent打了量化数据:Cognition减82%工作量,FactoryAI能看5000万行代码,值得一读。
Kepler AI将大型BI需求缩减为一个Slack提醒。Cognition测量到目标工作减少82%,而非token使用量。FactoryAI评估3000万至5000万行代码的代码库的代理就绪度。tryramp使用代理将客户输入转化为规格说明。
FDE is what happens when software gets easier to build but customer outcomes are still hard to come ...
FDE is what happens when software gets easier to build but customer outcomes are still hard to come by: - @kepler_ai_hq scoped a giant BI ask down to one Slack alert by Monday - @cognition measures an 82% reduction in targeted work, not token usage - @FactoryAI scores 30–50M-line codebases for agent readiness - @tryramp uses an agent to turn customer intake into specs Watch the full track: youtube.com/watch?v=1OMHGs… 💬 2 🔄 1 ❤️ 1 👀 391 📊 3 ⚡
- shao__meng07-27 06:31原文