Without knowing about this node, I wrote Re: Re: exactly 390 bytes.

It is not a complete obfu-engine, just lists compact encoding of the first few numbers. For some punctation chars, it gives a bit shorter encoding than yours (while it has longer ones for most others). For example:

_ = chr log sqrt exp hex be = chr oct hex ord uc yz
[download]

From these encodings, I especially like those relying on that oct stops scanning a number when it meets a digit 9.

What I find really strange is that expressions with length aaaaaaaaaaaaaa or those with length qr taaaaaat never give an optimal encoding.