Hi and welcome to this forum,
Pd's [random] is a pseudorandom number generator. This means that although the output "looks" random, it's actually being generated by a deterministic formula of which the results resemble a uniform distribution (the reason is why they "look" random). That said, if you start Pd and generate some random numbers, then quit Pd and do the same again, you will get the same results as before (due to the deterministic aspect mentioned before). To avoid this, you can send a seed to the number generator. If the seed is different each time, then the random numbers will also be different at each initialization. One of the tricks to get a unique seed every time is to use the CPU time to create a seed. See:
Also, if you are looking for a random number generator that always output different numbers on startup and does not need seeding, check miwi's objects here, his [rand] is excellent: https://github.com/0miwi0/pd-externals
Cheers,
Gilberto