On Mon, Oct 19, 2026 at 09:36:48PM +0300, Jarkko Sakkinen wrote: > Yeah, this is also a theme, as generally the tone of kernel development > has changed to be slightly more intrusive than before in the sense that > addressing architectural issue is often better strategy than doing > minimal effort as per code changes. > > This is because it is the way we actually decrease the LLM bug fix > activity. Implementation mistakes caused by the more bullish approach > has factors lower price tag, if it can affect the rate and density > of LLM storms. > > This is why e.g., I've put out a patch set that departs assoc_array > from kernel, and replaces it with rhashtable. This way we destroy > nests instead of mosquitos.
Within limits and not over the top of course but like if something is not working out for you look at one factor more generic problem and solve that. This way many quality issues get gradually addressed and it is actual antidote for LLM infections :-) > > > > > Surendran > > Br, Jarkko Br, Jarkko
