Increasing the prefix length again we reach `/26`, shown in white.
Now each of the numbers in the third byte breaks down into four address
ranges, corresponding to 0-63, 64-127, 128-191 and 192-255 in the fourth
byte. Now we have blocks of 64, on boundaries of multiples of 64.

**To summarize:** the IP address is a continuous string of 32 bits. Because
it is written as four 8-bit decimal values, it can be confusing to
visualize how various address prefixes fit together. Remember, byte
boundary or not, every time you increase the prefix length by one and move
the boundary one bit to the
right, the number of matching blocks is doubled, but the size of each
block is reduced by half. Conversely, decreasing the prefix length by
one and moving the boundary one bit to the left groups pairs of
blocks together, halves the total number of blocks, and doubles
the size of each. Always start at the closest byte
boundary and work the bits from there.

**Exercise:** I've illustrated the region around
`/24`, the boundary between the third and fourth bytes,
but it's important to remember that this diagram could be extended
downward, to the left and to the right almost indefinitely.
I suggest you try it for yourself - draw a diagram similar to
this one, only illustrating the region around `/16`, the
boundary between the second and third bytes. Mark the prefix blocks
corresponding to `/15`, `/14` and `/13`, then
do `/17` and `/18`. Pick some sample numbers,
construct IP address prefixes, and determine which sets of addresses
match.