Consider the five known primitive parity cycles
0,\quad 1,\quad 01,\quad 011,\quad 11110111000.
Call a binary word admissible (A) if it occurs as a contiguous factor in an infinite repetition of one of these cycles (allowing all cyclic rotations), and forbidden (B) otherwise.
An interesting observation is that at length 6 there are exactly 18 admissible words:
000000,\ 111111,
010101,\ 101010,
011011,\ 110110,\ 101101,
and the 11 length-6 windows of (11110111000)^\infty:
111101,\ 111011,\ 110111,\ 101110,\ 011100,\ 111000,\ 110001,\ 100011,\ 000111,\ 001111,\ 011110.
These 18 words correspond exactly to the
1+1+2+3+11=18
possible phases of the five known primitive cycles.
Moreover, length 6 distinguishes all 18 phases. Therefore, once an admissible 6-bit word is known, its continuation is uniquely determined. Hence there are still exactly 18 admissible words at every length k\ge6, out of 2^k possible binary words.
This suggests the following local formulation of the weak Collatz conjecture:
Conjecture. Every integer Collatz cycle has only admissible length-6 parity factors.
Equivalently, no integer cycle contains a B-word of length 6.
If this conjecture holds, every integer cycle must follow one of the 18 periodic rays determined by these length-6 words, and therefore must be a repetition/rotation of one of the five known primitive parity cycles.
So the global statement “there are no other integer cycles” can apparently be reformulated as a remarkably small local obstruction problem:
\boxed{\text{No integer Collatz cycle can contain a forbidden 6-bit parity word.}}
Note that this says that N cannot itself belong to a cycle if its first six parity bits form a B-word. It does not say that N cannot eventually enter a known cycle: a transient trajectory may contain arbitrary B-words before reaching it.
This raises an interesting question: can the impossibility of B-words inside an integer cycle be attacked directly through the associated numerator \Phi(w) and denominator 2^\ell-3^x?