If you are running workload that matters from infra footprint perspective on #Java , then should have planned upgrade to Java25 6 months ago. We did the upgrade to 25 from 21 on 2 large footprint services. Throughput is up ~10%, very conservatively. #Java25 1/2

The apache fory seems to be on a roll! The JSON benchmarks just look unrealistic!! We are running our own benchmark, but the numbers look too good to be true! #apachefory #java #performance github.com/apache/fory/...

I hope every day you all are trying to cultivate joy. A kindness to a neighbor. A long walk. Poetry. Birdwatching. Meditation. Being silly and playing peekaboo in a checkout line with a cute baby. These are dark times, and joy is something the bastards cannot take from us. Fight for it.

When fighting with maven, claude is generally relentless and is able to offload so much pain! #java #claude #maven

Great read surfingcomplexity.blog/2026/02/08/n...

Nobody knows how the whole system works

One of the surprising (at least to me) consequences of the fall of Twitter is the rise of LinkedIn as a social media site. I saw some interesting posts I wanted to call attention to: First, Simon W…

surfingcomplexity.blog

ah yes...I should also post about OUR book. the 2nd edition of "Observability Engineering" is now available for download.. and you can download it for free on the honeycomb site. this was supposed to be an update, and ended up being a total rewrite (& then some). 😅 www.honeycomb.io/blog/the-sec...

The Second Edition of Observability Engineering Is Here

IT’S HERE it’s here it’s here it’s here!!!! The second edition of Observability Engineering is available for download, and since Honeycomb is the sponsor, you can now download it from our website (th...

honeycomb.io

Doug Winter@doug.winter.cx · 2mo ago

Both Cat and Charity are very much worth reading.

Ran into an interesting thread starvation problem. After some digging, figured out that we had badly written code (when was it not). The take away is simple. Do not block indefinitely on future. DO NOT OVERENGINEER. Async is not easy. #Java

The LLM generated issue summaries/comparison look very bland and I tend to lose interest reading the same. Does anyone else feel the same too? The generated context generally over indexes on a particular aspect. May be I need getting used to, but currently its not helping. #AI

We've been working on a major schema migration at @polarsignals.com that has unlocked some pretty wild improvements. Apart from a 50% (!!) reduction in storage size thanks to logical compression in vortex.dev, there has been a dramatic improvement in query performance.

Love Caffeine cache. IMO, it just makes the cut as defacto in process cache impl. However, we only recently learned that for caches which needs to perform evictions (for eg: max size), cache initialized with .executor(Runnable::run) performs much better! #java #cache

AGAINST ALL ODDS, THOSE WHO BELIEVED BLUESKY THAT THE CRASH WAS DUE TO A DDOS AND THOSE WHO DIDN'T BELIEVE BLUESKY AND WERE SURE IT WAS BECAUSE OF THE AI VIBE CODING WERE BOTH RIGHT

Pookleblinky@pookleblinky.bsky.social · 4mo ago

pckt.blog/b/jcalabro/a... The postmortem of the *last* big outage shows that it was because they DOS'd themselves through vibecoded slop.

jlibdeflate, the new java JNI bindings to libdeflate, is pretty neat. Nice work by @tfenne.bsky.social t.co/AmzoFGZheT

Another great refresher of new capabilities of async profiler by none other than Andrei Pangin! Watch this if you are serious about profiling!! #performance #profiling #java youtu.be/u7-S-Hn-7Do?...

Advanced performance analysis with async-profiler by Andrei Pangin

YouTube video by Devoxx UK

youtu.be

Just got a 157m read draft HLD. AI is thriving!

10/ You can solve an enormous number of technical problems by taking a walk

5/ In general, work that shortens feedback loops is the most valuable "metaproject" work that can be done on a project. Doesn't matter what feedback loop: testing, compiling, deploying, getting customer feedback, analyzing data, whatever

TIL: Temurin build of openjdk always bundles the zlib implementation(since jdk21). This means you cannot use LD_LIBRARY_PATH or LD_PRELOAD to use alternate impl like the cloudflare fork, or zlib-ng. github.com/adoptium/tem... @adoptium.net #jdk #java

Watching the tech industry unleash unpredictable software into production in the name of AI adoption, then spend twice as much time and money trying to limit the blast radius, proves humans should be investing in our own intelligence before allowing the machines to mimic us.

A voice of reason. mariozechner.at/posts/2026-0...

Thoughts on slowing the fuck down

Thoughts on slowing the fuck down

mariozechner.at

Day <stopped counting> of using tcpdump to debug an issue in production!

Interesting morning! #copilot

Very early prototype. Thought of giving the java-ffi a run. Took the zstd-jni as comparison and asked copilot to generate the ffi version with as minimal garbage as possible. In the first iteration ffi is slightly better than jni! Time to dig into generated code. @java #java #ffi

@netflixeng.bsky.social wrote about Java Vector usage and how it transforms the ML inferencing landscape via netflixtechblog.com/optimizing-r... We have been doing the same for over an year with great success. We wrote about it at technology.inmobi.com/articles/202... #Java

Optimizing Recommendation Systems with JDK’s Vector API

By Harshad Sane

netflixtechblog.com

AI writing is like store-bought cake. It might be perfectly fine, maybe even as good as something you could make yourself, but it’s weird to give it to someone and say it’s homemade

We do some large scale NN inferencing on CPU. For us latency is paramount (sub ms). We have custom #java impl leveraging incubating vector. We ran benchmarks against Triton & ORT. Java impl is around 3 to 4x faster. One may say we haven't optimized the ORT/Triton well 1/2

if you love software and are feeling down about the state of software, find places where people are talking about building stuff, and most importantly, actually building stuff and talking about the specifics of the code

[new blog post] Butlers or Architects? muratbuffalo.blogspot.com/2026/02/butl...

Michiel Leenaars in his talk at #fosdem: "There is no Claude, just other people's code" (reference: there is no Cloud, just someone else's computer) @fosdem.org

My colleague Sanjay Ghemawat & I have done a fair bit of performance tuning of various pieces of code. We wrote an internal Performance Hints document ~2 years ago as a way of identifying some general principles & we've recently published a version of it externally. Doc: abseil.io/fast/hints.h...

DRAM prices mooning will hurt everyone and might cause a consumer tech recession on its own in the next 12 months even without the AI bubble popping. Hyperscalers are going to jack up prices for cloud resources, electronics manufacturers (Apple, Samsung, etc.) will raise prices, this is gonna hurt.