# [Python 3.8], <s>52</s> 50 bytes

-2 bytes inspired by [EasyasPi's answer][1].

Produces some integers with probability \$\frac 2 {256}\$ and some with probability \$\frac 3 {256}\$ in each iteration. Output includes a trailing `.0` for each number.

<!-- language-all: lang-python -->

 import os
 while id:print(id:=os.urandom(1)[0]%100)

[Try it online!][TIO-kkhgve2q]

[Python 3.8 (pre-release)]: https://docs.python.org/3.8/
[TIO-kkhgve2q]: https://tio.run/##K6gsycjPM7YoKPr/PzO3IL@oRCG/mKs8IzMnVSEzxaqgKDOvRAPIsM0v1istSsxLyc/VMNSMNohVNTQw0Pz/HwA "Python 3.8 (pre-release) – Try It Online"

Uses the builtin function `id` to avoid assigning a new variable before the loop. 
[`os.urandom(size)`][2] returns a `bytes` object with `size` random bytes. The `bytes` object behaves quite similar to a list of integers, which means `os.urandom(1)[0]` gives a single random integer from \$[0,255]\$, which we map to an integer from \$[0,99]\$ with a modulo operation.

---

# [Python 3.8], 53 bytes

Generates integers from a uniform distribution over \$[0, 99]\$.

<!-- language-all: lang-python -->

 from random import*
 while id:print(id:=randint(0,99))

[Try it online!][TIO-kkh0wocv]

[Python 3.8]: https://docs.python.org/3.8/
[TIO-kkh0wocv]: https://tio.run/##K6gsycjPM7YoKPr/P60oP1ehKDEvBUhl5hbkF5VocZVnZOakKmSmWBUUZeaVaAAZtiAVILaBjqWlpub//wA "Python 3.8 (pre-release) – Try It Online"


 [1]: https://codegolf.stackexchange.com/a/218184/64121
 [2]: https://docs.python.org/3/library/os.html#os.urandom