- JVM timing options
-
For as long as I have been coding in Java, we have had requirements to measure the execution time of blocks of code. While the current good practice is to use OpenTelemetry’s traces, not every company has reached this stage yet. Plus, some of the alternatives are OpenTelemetry-compatible. Let’s see them in order.
- TinyLFU: A Highly Efficient Cache Admission Policy
-
This paper proposes to use a frequency based cache admission policy in order to boost the effectiveness of caches subject to skewed access distributions. Given a newly accessed item and an eviction candidate from the cache, our scheme decides, based on the recent access history, whether it is worth admitting the new item into the cache at the expense of the eviction candidate.
I have learned about caches a bit when I worked for Hazelcast, but I never heard about TinyLFU. Thanks to Stefano Fago for the post on LinkedIn.
- WhatsApp’s Hidden Notes Told AI Not to Recommend a Regulator
-
IANAL, so I can’t judge whether it’s legal or not, and where. However, this is ethically so wrong on many levels. I’m happy I deleted my What’s App account when Facebook took over! Not to mention the possible siphoning of my phone’s data.
- Make yourself replaceable
-
This one is funny because I’ve always worked in this way, even though I understood the consequences for my job. I never thought about its benefits, though.
- Chinese AI Writes Weaker Code for Beijing’s Adversaries. We Fixed It. A Qwen3-Coder Case Study
-
So, you thought about using a so-called Open Source Chinese model to avoid paying the big players huge sums of money? Think again.
I tried to play with a Qwen model a couple of months ago, and asked why Xi was portrayed as Winnie the Pooh. It didn’t answer and accused me of being unrespectful.
- Setting up SSH keys on macOS with Touch ID
-
I so much want to try this!
- The Algorithmic Self-Portrait: Deconstructing Memory in ChatGPT
-
Our analyses reveal three key findings:
- A striking 96% of memories in our dataset are created unilaterally by the conversational system, potentially shifting agency away from the user;
- Memories, in our dataset, contain a rich mix of GDPR-defined personal data (in 28% memories) along with psychological insights about participants (in 52% memories);
- and ~A significant majority of the memories (84%) are directly grounded in user context, indicating faithful representation of the conversations.
Le Chat from Mistral AI also has this feature. It’s opt-in, and you can definitely fiddle with it.
- Anthropic Just Revealed 12 New Rules for Prompting Claude Opus 5.5
-
- Define What "Done" Means
- Stop Saying "Think Harder"
- Start With
mediumEffort - Give Long-Running Agents a Persistent Task List
- Tell the Agent When It Should Stop
- Don’t Assume
end_turnMeans "Task Complete" - Use Subagents Explicitly for Large Codebases
- Make Progress Updates Visible
- Explore Context Before Acting
- Treat Pasted Text as Untrusted Content
- Stop Retyping Information From Screenshots
- Don’t Say "Avoid Generic AI Design", Be Specific
- Bringing uv and ty together
-
I tend to use
uva lot, but haven’t usedty, and even less so since I let the coding assistants create the code for me. I should probably, types are one of the first guardrails against LLMs messing up the codebase. - Developer behind the AI-Built Adobe Clones Says They Are "Not Anywhere Close to Ready"
-
So the excitement, including mine, was a bit premature. And yet, months of tokens for a Photoshop clone is close to nothing.
- Don’t Be a Proxy Between Me and Your AI
-
I humbly admit I did it once. I was exposed. I don’t do it anymore.
- Python’s experimental JIT to be removed
-
One step forward, two steps backwards. I guess that if you aren’t a Pythonista following the developments, it may not be fully understandable.
A Java Geek weekly 160
JVM timing options. TinyLFU: A Highly Efficient Cache Admission Policy. WhatsApp’s Hidden Notes Told AI Not to Recommend a Regulator. Make yourself replaceable. Chinese AI Writes Weaker Code for Beijing’s Adversaries. We Fixed It. A Qwen3-Coder Case Study. Setting up SSH keys on macOS with Touch ID. The Algorithmic Self-Portrait: Deconstructing Memory in ChatGPT. Anthropic Just Revealed 12 New Rules for Prompting Claude Opus 5.5. Bringing uv and ty together.
