Egregoros

@clacke it's all bs anyway, someone did a statistical analysis of the bugs and found that it was just correlative to finding new bugs historically in rsync

@clacke https://alexispurslane.github.io/rsync-analysis/

@sun Thank you, that's interesting background info as I dive in.

The scripts used to fetch the data, collate it into a DuckDB database file, construct the views on that DB, and then do the statistical analysis on that data, were indeed written by GLM 5.1, as was the HTML and much of the original prose for the final report webpage you're looking at right now.

This is the first time I hear about GLM. So China (Z.AI) has an LLM that has a Claude-like profile, and offers a lightweight offline model.

@clacke china is producing multiple top-tier models you can run locally.

I think I can mostly predict what the camps will be and what they will say, but I'll still be interested to read it and learn something about where I stand.