Python Hub Weekly Digest for 2026-02-08
This week in Python news, popular topics included the tutorial on building pipeline parallelism from scratch, the creation of a game engine for Django, and the transition from Python 3.3's subprocess polling to event-driven waiting. In articles, notable pieces discussed speeding up Pillow's open and save, the introduction of asyncio compatibility for the Peewee ORM, and the performance trap of Django's time-based lookups. Interesting projects included OpenHands, an AI-Driven Development platform, and Kimi Code CLI, an AI agent for software development tasks. Have a great week and happy coding!
đź’– Most Popular
Al is killing programming and the Python community
OpenHands / OpenHands
🙌 OpenHands: AI-Driven Development
Let's Build Pipeline Parallelism from Scratch
The tutorial walks through building pipeline parallelism from the ground up, explaining how to split large AI models and training workloads across multiple GPUs to improve training efficiency. It breaks down concepts with step-by-step examples so developers can understand how data and compute are partitioned and coordinated in a distributed training system.
I created a game engine for Django?
The author built a multiplayer Snake game in the browser using only Python and Django LiveView, with no custom JavaScript, by keeping game state on the server and broadcasting rendered HTML over WebSockets.
From Python 3.3 to today: ending 15 years of subprocess polling
The post explains how Python’s long-standing busy-loop approach to waiting on child processes has finally been replaced with true event-driven waiting using OS primitives like pidfd_open() with poll() on Linux and kqueue() on BSD/macOS. This change eliminates constant polling, reduces CPU wakeups, and has now been adopted both in psutil and upstream into Python’s standard subprocess modu...
đź“– Articles
Oban, the job processing framework from Elixir, has come to Python
Speeding up Pillow's open and save
The author benchmarks and improves Pillow’s image open/save performance by avoiding unnecessary plugin imports and using lazy loading, leading to large speed gains in Python. Results show opening PNG images can be around 2.6 times faster and WebP up to 14 times faster, with similar improvements in saving images, and the changes are included in upcoming Pillow releases.
Stelvio – Ship Python to AWS
Xenia – A monospaced font built with a custom Python engine
Asyncio Finally Got Peewee
Charles Leifer introduces an asyncio compatibility layer for the Peewee ORM that uses greenlet to bridge synchronous database logic with asynchronous event loops without requiring a full codebase rewrite. This implementation allows developers to run standard Peewee queries and transactions within async contexts by using a task-local state management system and specific execution helpers ...
Django Time-Based Lookups: A Performance Trap
The article shows how certain Django time-based field lookups can bypass database indexes and turn a simple filter into a full table scan, dramatically slowing queries. It then demonstrates a much faster pattern by computing time boundaries in Python so the database can use the existing index and drop execution time from tens of seconds to under a second.
ruvnet / wifi-densepose
Production-ready implementation of InvisPose - a revolutionary WiFi-based dense human pose estimation system that enables real-time full-body tracking through walls using commodity mesh routers
PicoFlow – a tiny DSL-style Python library for LLM agent workflows
openai / skills
Skills Catalog for Codex
Engineering VP Josh Clemm on how we use knowledge graphs, MCP, and DSPy in Dash
Dropbox’s VP of Engineering explains how the company built Dropbox Dash, an AI-driven cross-app search and knowledge system that uses indexing, knowledge graphs, and contextual reasoning to unify work content from many tools into a single, context-aware platform. The piece highlights engineering decisions about when to use indexed retrieval over federated approaches, how MCP tool calls a...
4 Pyrefly Type Narrowing Patterns that make Type Checking more Intuitive
Learn about 4 ways Pyrefly narrows types, reducing the need to explicitly cast in your code.
Learn Matplotlib in 30 Minutes
In this video, you'll learn how to use the Matplotlib library in Python. If you're interested at all in data science, AI, machine learning, or scientific computing, then Matplotlib is a must learn.
Getting a custom PyTorch LLM onto the Hugging Face Hub (Transformers: AutoModel, pipeline, and Trainer)
A worked example of packaging a from-scratch GPT-2-style model for the Hugging Face Hub so it loads via from_pretrained, runs with pipeline, and trains with Trainer -- with notes on tokeniser gotchas.
Prototyping a Live Product Recommender With Python
The article shows how to build a real-time product recommender prototype in Python using Contextual Multi-Armed Bandits to simulate user behavior and validate online learning algorithms like LinUCB. It explains why bandits handle cold-start and context better than traditional models, walks through data generation, feature engineering, offline evaluation, and sets up a live simulation as ...
RAG for Legacy Systems: 7,432 Pages to 3s Answers
Production RAG for legacy systems: model-agnostic reranking validated across four LLM families. Real metrics, no vendor lock-in, 7,432 pages to 3s queries.
Python Hub Weekly Digest for 2026-02-01
redress
redress is a failure-policy library for Python services. It treats retries, circuit breakers, and stop conditions as coordinated responses to classified failure, rather than independent wrappers.
⚙️ Projects
kimi-cli
Kimi Code CLI is an AI agent that runs in the terminal, helping you complete software development tasks and terminal operations. It can read and edit code, execute shell commands, search and fetch web pages, and autonomously plan and adjust actions during execution.
Microcode
Microcode is an efficient terminal-based AI agent with an internal REPL environment for coding assistance. It leverages Reasoning Language Models (RLMs) to help developers with coding tasks directly from the command line.
TTT-Discover
Learning to Discover at Test Time.
Webrockets
High-performance rust powered websocket server for Python.
Polymcp
Polymcp provides a simple and efficient way to interact with MCP servers using custom agents.
Vision-Agents
Open Vision Agents by Stream. Build Vision Agents quickly with any model or video provider. Uses Stream's edge network for ultra-low latency.
earth2studio
Open-source deep-learning framework for exploring, building and deploying AI weather/climate workflows.
AutoMem
AutoMem is a graph-vector memory service that gives AI assistants durable, relational memory.
webctl
Browser automation for AI agents and humans, built on the command line.
dj-celery-panel
Your Celery task inspector inside the Django admin.
MimicKit
A lightweight suite of motion imitation methods for training controllers.
x-algorithm
Algorithm powering the For You feed on X.
hive
Outcome driven agent development framework that evolves.
bzfs
bzfs is a reliable near real-time, parallel replication and backup command-line tool for ZFS.
Project by Ruslan Keba. Since 2012. Powered by Python. Made in 🇺🇦Ukraine.