Cognition的人说写代码只占20%，真正难的是测试和部署。他们用Devin给大银行省了150%人力，值得听。
在aiDotEngineer会议上，Cognition的FDE负责人Jia Wu表示编码本身已基本解决，写代码只占20%的问题。真正的挑战在于测试、审查、部署以及企业级维护。该团队将Devin从“绝望时才会用”转变为为全球最大银行带来150%的额外人力。17分钟的演讲聚焦于AI演示与AI部署之间的差距，并指出多数编程工具都倒在这一环节。
*Jia Wu, @j_wu8 FDE lead at @cognition
*Jia Wu, @j_wu8 FDE lead at @cognition Limestone Digital @LimestoneHQ Deployed engineering lead at Cognition, Gia: "Coding itself is a mostly solved problem. Writing code faster - that's only 20% of the problem. The real problem is how do you test this code, review it, deploy it, and maintain it across the enterprise." One of the most honest talks at @aiDotEngineer on AI engineering this year. 17 minutes of pure signal from the team that took Devin from "I'd only use this if I was desperate" to delivering 150% extra headcount at the world's largest banks. The gap between AI demo and AI deployment is where every coding tool dies. This team closed it. If you're still measuring AI productivity by how fast you write code, you're solving the wrong 20%. Watch it, then read the article on AI-native engineering at enterprise scale below. Your browser does not support the video tag. 🔗 View on Twitter 🔗 View Quoted Tweet 💬 0 🔄 0 ❤️ 1 👀 645 ⚡