JFrog's security team uncovered a PyPI admin key leak that could've been catastrophic - we're talking access to 200+ repositories across python, pypa, psf, and pypi organizations! The admin had hardcoded the key in source code and later removed it, but it lingered in cached pyc files. This shows that source code security scanning isn't enough - binary security scanning is equally crucial!
TL;DR
A leaked PyPI admin token nearly caused the worst supply-chain attack imaginable, caught by binary secret scanning; free-threaded CPython opens up for experiments, and exo turns everyday home devices into an AI cluster.
Articles & Tutorials
× 12The PSF Infrastructure Director (the person involved in the previous incident) published an incident report revealing that the leaked key had been committed about a year ago - thankfully, it wasn't maliciously exploited! Key takeaways: set appropriate expiration times for tokens, exclude __pycache__ and *.pyc files when building Docker images, and keep your source code clean.
From the "Learning Security Testing from 0 to 1" series, this article explores Python prototype chain exploitation and pollution through a Flask web app's server-side template injection attack. It helps readers understand prototype chain concepts, mechanisms, and potential security risks in Python.
Python 3.13 is dropping in just over 2 months - ready to experiment with the GIL-free feature? This article introduces the free-threaded-compatibility repository, provides resources, and tracks open-source project support for no-GIL. It also covers two thread safety issues discovered in numpy and pywavelets.
Learn how to use the profilehooks library to detect Python code performance, then leverage Snakeviz and gprof2dot visualization tools to analyze data, identify major performance bottlenecks, and achieve effective optimization.
Pastebin is an online service for sharing code and text. This article shows how to build one with Flask, supporting text pasting, programming language selection, and shareable URL generation, with detailed explanations of each code component. (Fun coincidence: we also received a long-awaited open-source project submission this week - a simple pastebin built with async Flask, shared by @yuxiaoy1!)
@Manjusaka strikes again with another CPython bug fix! In Python 3.13 Beta, the PEP-683 implementation plus surrounding changes caused some extensions compiled under older versions to fail in Python 3.13.
@Komeiji Satori also fixed a CPython bug! It was about Python's tuple cache pool issue - the final code change was just adding an "=" sign, but the process of discovering and reproducing the issue from source code really deepens your understanding of the Python interpreter.
What should you watch out for when writing Python decorators? The article's answer: always use functools.wraps - without it, multiple decorator usage can cause naming conflicts.
The article argues that using requirements.txt, manually creating and entering venv virtual environments, manually setting PYTHONPATH, and using setup.py to build extensions are all bad practices. The author presents a poetry-based project management solution instead.
Ever wanted to build your own GPT? Curious about how large models are trained? This article introduces and implements the basic components of large models, like tokenizers and data loaders. (Note: Part 2 of this series is Building and Training GPT-2)
Our newsletter's Season 1 and Season 2 each featured around 800 links and over 60,000 words - that's pretty mind-blowing! Wonder how many people have actually read everything? This article extracts all Season 1 links, categorized by type, making it easy for newcomers to browse and discover gems they might have missed!
Projects & Resources
× 12Build AI clusters on your iPhone, iPad, Android, Mac, and Linux devices. Supports LLaMA and other popular large models with automatic cluster device discovery. (2K stars)
A sleek text user interface built with Textualize for managing and consuming Kafka topics, with Protobuf deserialization support.
A simple visual web scraping project built with Requests + BeautifulSoup4 + PyQt5. Supports scraping multiple elements from multiple URLs, handles login/registration, and features comprehensive logging.
Runs scrapers periodically to get information feeds and updates corresponding RSS sources, or saves data and generates RSS feeds when APIs receive data. (Shared by @AlterNao1)
Supports automatic saving, querying, and version control of Python function-level operations, with the ability to store and track individual elements in collections.
Lets you update code while it's running! When you modify functions, it affects all call sites simultaneously; when you modify modules, it only reruns the changed lines without loading the entire module. (1.1K stars)
A production-oriented full-stack solution that achieves SOTA results on many public speech datasets. (3.9K stars)
Provides an intelligent, self-improving memory layer for large models, enabling personalized AI experiences across platforms and applications. (10.8K stars)
Features Magic-PDF and Magic-Doc capabilities, converting PDFs, web pages, or multi-format e-books to Markdown format.
A prompt template management project built with Flask + MongoDB, supporting basic CRUD operations.
Contains nearly 4,000 disposable domain email addresses, useful for registering virtual users. (2.7K stars)
A framework specifically designed for generating embeddings, supporting popular text models. No GPU or PyTorch dependencies required - uses ONNX runtime and outperforms OpenAI Ada-002. (1.2K stars)
Podcasts & Videos
× 2From the podcast hosted by IEEE Computer Society and IEEE Software magazine, featuring a senior software engineer at Pydantic who's also a maintainer of Starlette, Uvicorn, and other projects. The discussion focuses on FastAPI's design, implementation, and applications.
Harlequin enables visual SQL operations in the command line terminal. This podcast episode features the library's author discussing the project's origins, why they chose TUI over GUI, lessons learned, and more.
Discussions & Questions
× 2A Reddit thread exploring the best/newest frontend tech stack options. HTMX appears with surprisingly high frequency!
A V2EX thread discussing various Python task management solutions.