“The G in GPU is for Graphics damnit!”: Adventures in Triton Kernels, Profiling, Parallelism and More

The blog explains how Triton, a Python-based DSL for CUDA programming now developed by OpenAI, enables writing custom GPU kernels with better performance and ease than direct CUDA. Through a case study simulating Physarum slime mold, the post showcases profiling, optimization, and how Triton can vastly outperform PyTorch by fusing operations and reducing memory overhead.

https://ut21.github.io/blog/triton.html

SINQ

A novel, fast and high-quality quantization method designed to make any Large Language Model smaller while preserving accuracy.

https://github.com/huawei-csl/SINQ

Introducing Air: A Pioneering AI-First Python Web Framework

The Air framework is a pioneering AI-first Python web framework built on FastAPI and Starlette, designed to optimize developer and AI agent productivity with features like intuitive Air Tags for HTML generation, seamless Jinja template integration, and built-in HTMX support. Inspired by frameworks like Django, Flask, and Pyramid, Air emphasizes modularity, modern developer experience, an...

https://audrey.feldroy.com/articles/2025-10-06-air-pioneering-ai-first-python-web-framework

Python 3.14's best new features

Python 3.14 introduces practical, user-friendly features such as syntax highlighting in the REPL and debugger, improved error messages with helpful suggestions, and pathlib enhancements like copy and move methods. It also adds template string literals (t-strings) for advanced string processing, deferred evaluation of annotations, new UUID versions sortable by creation time, and better co...

https://www.pythonmorsels.com/python314/

gemimg

Lightweight wrapper for generating and editing images from Gemini 2.5 Flash Image/Nano Banana.

https://github.com/minimaxir/gemimg

Functions: a complete reference

This article serves as a complete reference for all the non-trivial things you should know about Python functions.

https://mathspp.com/blog/pydonts/functions-a-complete-reference

FuncToWeb

Transform any Python function into a web interface automatically.

https://github.com/offerrall/FuncToWeb

T-Strings: Do They Actually Make SQL Better in Python 3.14?

The video explains that t-strings are a new string format in Python 3.14 designed to help with SQL query sanitization by separating query parts and parameters for better readability and safety. While t-strings can improve managing complex queries with many parameters, for simple queries they add unnecessary complexity compared to traditional parameterized queries, making them useful main...

https://www.youtube.com/watch?v=R5ov9SbLaYc

Why Reactive Programming Hasn't Taken Off in Python (And How Signals Can Change That)

Reactive programming offers significant benefits for Python applications - it reduces bugs, simplifies complexity, and improves maintainability. Yet most Python developers avoid it. The problem isn't reactive programming itself, it's how we've been doing it. Python's reaktiv makes reactive programming as simple as spreadsheet formulas.

https://bui.app/why-reactive-programming-hasnt-taken-off-in-python-and-how-signals-can-change-that/

Training a Deep Learning Model for Echogram Semantic Segmentation

In this tutorial we build a deep-learning pipeline for echogram segmentation using open-source tools. Echograms are two-dimensional plots of acoustic echo intensity versus time and depth recorded using sonar instruments, in our case echosounders.

https://oceanstream.io/training-a-deep-learning-model-for-echogram-semantic-segmentation/

Python 3.14 Is Here. How Fast Is It?

Python 3.14 delivers notable performance improvements, including up to 27% speedup in some benchmarks compared to Python 3.13, and enhanced free-threaded execution for better multithreading. The new release also adds features like syntax highlighting in the REPL and improved concurrency support.

https://blog.miguelgrinberg.com/post/python-3-14-is-here-how-fast-is-it

How to Level Up Your Python Logs with Structlog

This Structlog guide covers configuration context JSON output error handling and OpenTelemetry integration to make your logs a useful signal for Observability.

https://www.dash0.com/guides/python-logging-with-structlog