JProfiler Get rid of your performance problems and memory leaks! Modern Garbage Collection Tuning Java Performance Training Courses COURSES AVAILABLE NOW. We can provide training courses to handle all your Java performance needs Java Performance Tuning, 2nd ed The classic and most comprehensive book on tuning Java Java Performance Tuning Newsletter Your source of Java performance news. Subscribe now! Enter email: Training online Threading Essentials course JProfiler Get rid of your performance problems and memory leaks!

Back to newsletter 304 contents

Like most dev, I'm using AI to code assist. Or rather trying, because even the latest models are only at the level of a junior coder, so for actually creating good code they are of limited usefulness. But we all know the actual coding is a small part of a dev's job. The AIs are helpful in many areas, like explaining unfamiliar code, planning, specifications, etc. I decided to apply the AI (trying with the three most popular models at their latest versions) to creating the newsletter. My own process is to scan around 100 relevant articles, talks, blogs, etc. Of these 100, I filter out the no-hopers and I'm left with around 30-40 that are potentially useful for the tools, news and tips sections. Then I go into more detail with these, ending up with 3 for each section. It's a process that I would have thought the AIs could handle well.

They were pretty bad at finding stuff. If it's very directed specific web searching, they're good, but for a more generic trawl of what might be of interest for the newsletter, worse than meh. They were also bad at filtering out no-hopers, basically because they were unable to apply good quality scoring. If I could codify how I decide on the quality and relevance of the article then they could do it, but if I could do that I could write a program to do it - and I can't. Finally they were also quite bad at choosing what to go into the newsletter from those last 40, until I repeatedly pointed out flaws in how they were scoring the content. Without a lot of feedback they chose some acceptable options but also some really bad options, and almost never the best options. With sufficient repeated feedback they were a lot better, but by then it had taken longer to work with any one AI than doing it all myself! Hopefully that was good experience for when the models get better.

Now on to all the manually built newsletter list of links, tips, tools, news and talks, and as usual I've extracted all the tips into this month's tips page.

A note from this newsletter's sponsor

News

Java performance tuning related news

• Automated Java performance analysis with Azure Performance Diagnostics tool
• jaz, a resource-aware optimization pipeline
• BellSoft hardened Java image use 30% less RAM and disk space

Tools

Java performance tuning related tools

• Eclipse OpenJ9 garbage collection policies and heap tuning
• good reference article on JVM internals lifecycles
• How to use the decorateBy operator for querying JFR events

Articles

• A Java Developer's Quest for IO Performance (David Vlijmincx/P99 CONF)
• Get The Most Out of Virtual Threads in Java 25 (Christian Worz/Devoxx)
• The Ultimate Guide to Java Garbage Collection Tuning (Annya Arun/GcEasy)

Jack Shirazi

Back to newsletter 304 contents