Claude Opus 4.7 is measurably better than Opus 4.6 for Bolt's longer-running app-building work, up to 10% better in the best cases, without the regressions we've come to expect from very agentic models.
在长时间应用构建中实现10%的提升且没有常见回归问题,这表明AI在持续任务执行上的稳定性取得了重大突破,'pushes the ceiling on what our users can ship in a single session'暗示了AI对软件开发范式的根本性改变。