[suby]

I feel like it should be possible to build a universal accessibility tool using machine learning that relies on vision rather than native framework UI/toolkit support. The problem seems much more tractable than other AI projects like self-driving cars, and also more in the realm of existing successes that we've already seen with machine learning. The problem of inaccessible apps will only get worse as wasm becomes more mature, I feel like the only way to win is to sidestep the problem.