Wow. This is some deep learning

No. To start, there is far more data needed than "matter" or "no matter", unless the only thing you are trying to model is gravity, and even then how much matter is vitally important.

Second, the universe doesn't actually consist of blocks. The planck length is the smallest length you can measure due to how light works, but that doesn't mean everything fits into integer multiples of it. And string theory suggests there are structures in the universe at even finer scales.

Thirdly, Quantum mechanics says that objects do not even have fully specified locations. They exist in a probabilistic manner, so there is no hard edge of "matter here" and "not matter there".

Fourth, it doesn't account for things which are not matter, like photons, which are extremely important.

Fifth, the mapping to translate an index into this string into a 3d coordinate is shifting. The universe is expanding, so the conversion to and from the byte stream is going to change moment to moment, and so the array is unworkable.

Sixth, moments aren't discrete timesteps either. It is continuous, so you can't just break it down to a series of simple steps you apply rules on and remain accurate.

Seventh, not only is time not discrete, its not even consistent. Relativity says time flows at different rates in different places and for different things.

Eight, the universe is not deterministic. I will grant that taking in the set of all previous states of the universe in their entirity does get around bell's inequality, it does mean that your model is non-local.
#Researched