Vincent Bonin

boninlab.bsky.social

20 October 2026 07:48

I am not a Python coder. A postdoc in my lab just thought me that 42 is from Hitchhiker's Guide to the Galaxy, a typical example for basic Python.

Vincent Bonin

boninlab.bsky.social

20 October 2026 07:34

top is step 1 (correlating the whole 400 sample sequence) which is resolved with a simple for loop, bottom is step 2. eye inspection of bottom trace shows which data segments to redo step 1 to estimate the missing parameters. had the authors not been so lazy it would have been impossible.

Vincent Bonin

boninlab.bsky.social

19 October 2026 17:34

One comment: the commenter did not have to explore the full code space to reproduce the data. They only had to (1) correlate the data with random sequences of different seeds to find the seed and (2) use the deviations from the sequence to constrain the generator for each subsequences.

Daniel A. Walker

danwalker9999.bsky.social

19 October 2026 08:40

This is more chilling than the commonly-exposed practice of faked images. If datasets are being churned out with tricks such as pseudo-random number generators, then it is possible that the contamination of science by fraud is even deeper than we thought. @sholtodavid.bsky.social

Daniel A. Walker

danwalker9999.bsky.social

19 October 2026 08:40

21 authors. 21 affiliated institutions. 4 countries (China, South Korea, USA, Australia).