Please RT! We need those papers to make #HaystackConf an awesome event - have you got a #search story to tell?

I'm planning to get off twitter by the end of the month. If you'd like to stay connected, please ping me.

If your an ASF committer definitely check out this event (and then give me a summary after since I have an offsite commitment that overlaps).

More dimensions isn't always better. Just ask yourself, is 10% more signal with increased hallucinations worth a 99% increase in storage cost? #GenAI #vectors #search #lucenia

Lucene's nightly benchmarks got a massive speedup on queries sorted by numeric field last night: people.apache.org/~mikemccand/lu…. This is due to this PR: github.com/apache/lucene/….

Community Over Code EU is kicking off today in Bratislava, Slovakia! For anyone who lives in/around Slovakia who is interested in attending, check out our 1-day pass for locals. eu.communityovercode.org/tickets/ #CommunityOverCode #opensource

#CommunityOverCode keynote: EU legislative changes panel: "Many of these standards depend on working with people, which is always hard" Code is easy. People are hard. Optimize for people.

Search track at Community Over Code in Bratislava, atitaarora kicks off the day talking about “Navigating Challenges and Enhancing Performance of LLM based Applications” #communityovercode #lucene #solr #theasf

In the follow up talk on the Search track, we have Alessandro Benedetti talk about “Hybrid Search with Apache Solr” #communityovercode #theasf #solr #lucene #llm #search

The talk by Angad Sharma talking about Solr at zomato #communityovercode #lucene #solr #zomato

Next up is Yupeng Fu from Uber talking about their use of Lucene for Vector search at Uber and Uber eats! #communityovercode #lucene #solr #vectorsearch

⁦Dirk-Wiǁem van Gulik⁩ explaining the reasons for the CRA: the Grover Shoe Factory explosion is one example. Fast moving industries move fast and break things - which harms society when steam boilers actually blow up and kills people. #CommunityOverCode

That picture says a lot! #communityovercode

If you want great AI, talk to a search person.

"Introduction to Information Retrieval " never gets old. (#53) nlp.stanford.edu/IR-book/inform…

Here’s the thing folks. I’ve been coding 32 years. When something like this happens it’s an organizational failure. Yes, some human wrote a bad line. Someone can “git blame” and point to a human and it’s awful. But it’s the testing, the Cl/CD, the A/B testing, the metered

Thank you Vigya Sharma for improving #Lucene's benchmark tooling specifically to make testing Lucene's KNN search a bit easier! Benchmark tooling doesn't get enough open source love when it is arguably more important than the software it is testing. It is a compass that helps

A one-line removal recently improved performance for one query in Lucene's nightly benchmarks by ~2x people.apache.org/~mikemccand/lu… (annotation GP) github.com/apache/lucene/… This is because this line prevented dynamic pruning from using impacts on the lowest skip level on high-freq terms.

Lucene just changed the way it stores skip data, from up to 10 levels and stored separately, to at most 2 levels and inlined in postings. This simplification helped speed up queries that botttleneck on advancing by small intervals, see annotation GS at people.apache.org/~mikemccand/lu….

Lucene 10 was released earlier today. The main release highlight is improved hardware efficiency, I wrote about it on the Elastic blog: elastic.co/search-labs/bl…