I'm running this test snippet using numpy 1.18.4 and am a bit confused about the output:

import numpy as np

print(np.version.full_version)

sseq = np.random.SeedSequence(42)
sseq_child = sseq.spawn(1)[0]

rng = np.random.default_rng(sseq)
rng_child = np.random.default_rng(sseq_child)

print(rng.random(10) - rng_child.random(10))

This outputs ten zeros, which surprises me. I would have expected that the first child sequence spawned from sseq would actually be different from sseq itself, but it appears to be in fact identical.

Is this intended? If so, I didn't see any mention of this in the documentation, and I'm a bit worried that people will actually expect (like I did) that random numbers generated from a SeedSequence object and its children would be totally independent.