- Choosing a dependency
-
In this post, I’d like to describe my approach when choosing a dependency. I’ll first define what I mean by dependency in the context of this post. Then, I’ll list a grid of several criteria to analyze possible dependencies with.
- Language Modeling from Scratch, a free course by Standford University
-
Language models serve as the cornerstone of modern natural language processing (NLP) applications and open up a new paradigm of having a single general purpose system address a range of downstream tasks. As the field of artificial intelligence (AI), machine learning (ML), and NLP continues to grow, possessing a deep understanding of language models becomes essential for scientists and engineers alike. This course is designed to provide students with a comprehensive understanding of language models by walking them through the entire process of developing their own. Drawing inspiration from operating systems courses that create an entire operating system from scratch, we will lead students through every aspect of language model creation, including data collection and cleaning for pre-training, transformer model construction, model training, and evaluation before deployment.
- Introducing Flagship: feature flags built for the age of AI
-
The AI part may be bait, but the rest is an interesting approach: set the feature flag at the edge.
- How to Never Hit Claude’s Usage Limit Again: 3 Proven Methods
-
- Method 1: Optimize How You Use Claude
- Method 2: Batch Your Requests and Use Projects
- Method 3: Optimize Your Prompts | Claude’s Batch API
- Hand‑drawn QR codes
-
What a fun idea!
- AI fatigue is real and nobody talks about it
-
The post resonates within me.
What actually helped:
- Time-boxing AI sessions
- Separating AI time from thinking time
- Accepting 70% from AI
- Being strategic about the hype cycle
- Logging where AI helps and where it doesn’t
- Not reviewing everything AI produces
- RIP Commercial OCR. An Open-Source Model Just Topped Every Benchmark.
-
I never needed OCR for my projects, but it can be quite useful. However, I found one of the comments to be a bit misguided.
But because it shows that in specialized domains, smaller focused models can still outperform general-purpose giants.
A properly-trained specialized model outperforms general-purpose ones on the task it was trained on. It’s as simple as that.
- Span Metrics Connector
-
Connectors act as receiver on one end and received on the other hand connecting two pipelines.
Aggregates Request, Error and Duration (R.E.D) OpenTelemetry metrics from span data.
- Your Onboarding Is a Hazing Ritual and You Call It Agile
-
I personally didn’t remember a single prepared onboarding. Most of them, I was told to read documentation first, without guidance. I never worked over the week ends to compensate for the lack of proper onboarding and I was never fired. But I work in Europe, where there’s no culture of overtime.
- Taming Maven Version Ranges
-
The post explains how, but I still don’t understand the why. Plus the syntax is abysmally complicated.
- Camoufox
-
- Built for AI agents
- Undetectable by design
- Fingerprint injection & rotation
- Anti Graphical fingerprinting
- Quality of life features
- Decomposing OpenTelemetry Collector Configuration for Maintainability
-
- Splitting by component type
- Splitting by signal pipeline
- Environment-specific overlays
A Java Geek weekly 136
Choosing a dependency. Introducing Flagship: feature flags built for the age of AI. How to Never Hit Claude’s Usage Limit Again: 3 Proven Methods. Hand‑drawn QR codes. AI fatigue is real and nobody talks about it. RIP Commercial OCR. An Open-Source Model Just Topped Every Benchmark.. Span Metrics Connector. Your Onboarding Is a Hazing Ritual and You Call It Agile. Taming Maven Version Ranges. Camoufox. Decomposing OpenTelemetry Collector Configuration for Maintainability.
