Krystian Safjan's Blog

So OpenAI deprecated gpt-4o-mini.

The illusion of control in AI-assisted engineering Dashboards are green.

1.

Overview Homebrew distribution lets users install your Python CLI with brew install your-tool.

In a Hacker News discussion on the article Management as AI Superpower by Ethan Mollick, I came across a comment that resonated strongly with a realization I’ve had recently while coding with a factory.ai droid: The limiting factor at work isn’t writing c…

Option 1: Static Data Approach (Recommended) Key Points Option 2: Simple predict_fn (No Tracing Decorator) Comparison: When to Use Each Approach Complete Working Example Recommendation You can use MLflow's RAGAS integration to calculate evaluation scores…

RAG Evaluation with RAGAS and MLflow Table of contents Why This Tutorial?

The Mental Model Git worktrees let a single repository sprout multiple working directories, each tied to a different branch but sharing the same object store.

A practical step-by-step tutorial using Git worktrees, small and large refactorings, and GitHub Copilot.

Understanding the Problem Homebrew is designed to install software without root privileges.

The Basic Idea Introduced in Python 3.8 via PEP 544, Protocols extend Python’s static typing system by allowing structural subtyping — also called static duck typing.

Understanding orphan branches in Git There’s a lesser-known Git feature that allows you to start a branch with no history at all: git checkout --orphan <branch>.

The Mental Model Python typing wasn’t born overnight.

The Mental Model: What Is the Language Server Protocol The Language Server Protocol (LSP) is one of those invisible technologies that quietly revolutionized the developer experience.

The Core Idea CSS variables (custom properties) let you store reusable values directly in CSS and update them at runtime.

Core Principle The * by itself in a function signature forces everything after it to be keyword-only arguments.

What makes Shape Up different?

The High-Level Approach Introduced in Python 3.2 via PEP 3148, concurrent.futures gives you a unified interface for running code in parallel.

Core Principle Python has two built-in ways to run code concurrently: threading and multiprocessing.

Core Principle async/await is Python's way of writing concurrent code that can handle multiple I/O operations without blocking.

I’ve always liked Make.

I've been diving into how to use the official openai Python package to talk to local Ollama models—and when it makes sense to bring in abstraction layers like LiteLLM.

Over the past few months, I’ve been using Jupyter notebooks to explore and document various data analysis tasks.

To downgrade your project’s virtual environment e.g. from Python 3.11 to 3.10 using uv, here’s a step‑by‑step process: 1.

Introduction The Coverage Trap Beyond Line Coverage: Behavioral Auditing The API-First Testing Strategy Automating the Tedious Parts The Flaky Test Problem Test Quality as a First-Class Concern Integration and End-to-End Validation Looking Forward Introdu…

When working with Python objects, understanding how to properly copy them is crucial for avoiding unexpected behaviors in your code.

When hunting for the origin of mysteriously defined by unknown plugin zsh aliases: zsh -xv 2>&1 | grep "alias_name" This works by: Starting zsh with -x (trace) and -v (verbose) flags Redirecting both stdout and stderr (2>&1) to capture all output Filteri…

I've been using VS Code for years, and keyboard shortcuts can definitely get messy, especially when you have multiple extensions installed.

NetworkX - The Python Swiss Army Knife RDFLib - When You Need Semantic Power PyGraphviz - Visualization with Query Capabilities DIY Solution - Custom Graph Structure Making the Right Choice Performance Considerations Beyond Simple Solutions A Note on Auto…

GitHub Actions workflows are powerful automation tools, but sometimes you need to temporarily disable them.