刘年丰：最本质的原因就是因为，我们现在具身模型主流使用的VLA，是沿袭的动态模型沿袭了大语言模型——对整张图片做全局信息映射。
这个框架类似苹果的 App Intents。在苹果的构思中，用户可以使唤 Siri 来操作各种 app 来实现功能，而底层实现方式就是通过 App Intents ——新一代 Siri 迟迟不能落地的前提下，App Intents 足以提供不错的效果。
。关于这个话题，91视频提供了深入分析
There is no syscall surface to attack because the code never makes syscalls. Memory safety is enforced by the runtime. The linear memory is bounds-checked, the call stack is inaccessible, and control flow is type-checked. Modern runtimes add guard pages and memory zeroing between instances.
– If the location or view is unrealistic or fictional, create a composition blending both the location and the view into a single scene.