That should have been "We threw out >99.9% of the results as nonviable".
The final result was 16 viable phages out of 700,000 sequences.
This is just random mutagenesis with extra steps.
Aug 10, 2026, 4:20 PM@michael_w_busch This could just as easily read " Viruses Generated with Bash Scripts" or "Viruses Generated By Excel Spreadsheets" 😒
@michael_w_busch I genuinely think someone with expertise on the topic could probably quickly think up a super simple model for screening sequence function, spend ten minutes writing a Python/R/etc script that generates sequences and filters them based on those back-of-the-envelope criteria, and do a better job than that. The delusional nature of the "AI" hype is really stunning.
There is an entire section of King et al. 2026 where the authors describe how they adjusted the probability distribution for the random sequence text generator, as well as detailing the filters they imposed to exclude all of the obviously nonviable results.
It includes this bit: "we filtered out generated sequences containing non-nucleotide characters".
Which would be impossible if they had simply written their own script rather than using the text generator.
So. Yeah.
Aug 10, 2026, 5:51 PM@michael_w_busch @lcwheeler Ha!
AACTATGAGGTC🤪GATTACATTGACCA
@michael_w_busch @wcbdata Simply stunning...