מדדים לאימון AI

תקציר מקורי באנגליתAn engineer at a large tech company built a voluntary dashboard showing everyone's AI token usage, and colleagues promptly started competing to top it. Mingsheng Hong uses that as the thing not to do. His team runs the same dashboards, but treats them as a smoke detector rather than a leaderboard: a team using surprisingly few tokens is worth a conversation, and nobody should ever be rewarded for burning more. He draws the parallel to lines of code, a number worth tracking and a terrible thing to optimize, given that deleting code is often the better outcome. The pitfall he flags is going straight from measuring cost to cutting it, because that is only one side of a ratio. So Ironclad measures value too, and the metric evolved in public: lines of code, then open pull requests, then merged