Nice! I think the "floating point common names" page could be extended though, namely because it doesn't mention the bit amounts, but only decimal meanings like "s23e8". I think it would be more helpful to say that half is 16-bit, float is 32-bit, and so on, especially since in some programming languages these are called f16, f32 and f64 or float16, float32 and float64.

I also think an early chapter could actually introduce what floating point is in the first place, now the first chapter about floating point is immediately about the detailed topic of compression.