Skip to content

Details

Dear all,

Software updates are a continuous task. Users keep asking the same question: what changed, and is it safe? This talk traces my research trying answering that question. We started with changelogs and dependency graphs. Then came machine-learned "breaking update" classifiers, enriched with Reddit sentiment analysis. Then a new model estimating n-to-n+1 update risk. Today, agentic AI systems reason over all of it and answer directly. In this meetup, we'll review what's working, what still fails, and some open questions about where it might go next.

Please bring your experience and question.

Team Antrunner.

You may also like