Technology Radar
Published : Nov 05, 2025
NOT ON THE CURRENT EDITION
This blip is not on the current edition of the Radar. If it was on one of the last few editions, it is likely that it is still relevant. If the blip is older, it might no longer be relevant and our assessment might be different today. Unfortunately, we simply don't have the bandwidth to continuously review blips from previous editions of the Radar.
Understand more
Nov 2025
Trial
代码迁移的形式多种多样，从语言重写到依赖项或框架升级，几乎都不是一件小事，往往需要数月的人力投入。我们有一个团队在升级 .NET 框架版本时，尝试用 AI 缩短这一流程。过去，我们曾推荐 OpenRewrite 这类确定性、基于规则的重构工具。单靠 AI 自动升级通常代价高昂，且易陷入无效对话。于是，团队把传统升级流水线与智能编码助手结合，推进复杂的转型。区别于全程交由 AI 升级，他们将流程细化为可验证的小步骤：分析编译错误、生成迁移差异、反复验证测试。这种混合方法让 AI 编码智能体成为软件维护中的实用协作者。业界案例，如 Google 的大规模 int32 到 int64 迁移，也展现出同样趋势。尽管这一实践在节省时间方面结果好坏参半，但减少开发者重复劳动的潜力十分明显，值得持续探索。
