The rate of AI progress.

What's required for full dive are molecular nanotechnology and brain computer interfaces, and while those areas of research aren't yet accelerating like AI, they will soon; I expect proto-AGI will be created in either 2023 or 2024 (probably because of the scaled up Gato), which should accelerate all lines of research faster than anything before, so that's how I came up with my time frame prediction of 2029 for full dive (and AGI in general).