- Python Weekly
- Posts
- Python Weekly (Issue 722 November 13 2025)
Python Weekly (Issue 722 November 13 2025)
Rahul Chaudhary
November 13, 2025
Welcome to issue 722 of Python Weekly. Let's get straight to the links this week.
Voice AI: Get the Proof. Avoid the Hype.
Deepgram interviewed 400 senior leaders on voice AI adoption: 97% already use it, 84% will increase budgets, yet only 21% are very satisfied with legacy agents. See where enterprises deploy human-like voice AI agents - customer service, task automation, order capture. Benchmark your roadmap against $100M peers for 2026 priorities.
News
PyCon US 2026 invites proposals for Talks, Tutorials, Charlas, and Posters through December 19, 2025, with submissions open to speakers of all backgrounds. The conference in Long Beach adds new tracks on The Future of AI with Python and Trailblazing Python Security, and all presentations will be delivered in person.
Articles, Tutorials and Talks
Rippling redesigned their Django monolith’s deployment model using Gunicorn’s pre-fork mode, enabling the master process to load the app and fork workers, which then share memory thanks to Linux’s Copy-on-Write. By systematically cleaning up threads and connections before forking and using memory optimizations like gc.freeze() and jemalloc, they achieved over 70% reduction in memory usage and a 30% decrease in compute costs, while also improving resilience and startup speed.
The article from Slack Engineering describes how Slack improved their build pipeline by applying classic software engineering principles combined with high-performance build tooling (Bazel). They restructured their build system by separating concerns, modeling builds as graphs of well-defined work units, and leveraging caching and parallelization, which resulted in builds running up to six times faster and enabling faster feedback cycles and more frequent releases.
Django Admin Deux is a modern, extensible replacement for Django’s default admin interface, featuring a factory pattern architecture and a robust plugin system. It aims to make customizing and extending the admin UI much easier, bringing new capabilities and flexibility to Django project administration.
The article explores how frequently Python allocates memory, focusing on the integer type implementation. It finds that CPython allocates new memory objects often but mitigates overhead by reusing objects from a freelist and employing a specialized pool allocator; nonetheless, Python's integer handling still incurs considerably more allocation overhead than the single CPU ADD instruction it theoretically should take.
The video explains what “Pythonic” code means, refactoring a messy script step-by-step to demonstrate idiomatic Python practices like using functions over classes, dataclasses for structure, context managers, type annotations, EAFP error handling, pathlib, logging, and organizing code in a main function. The core message is that Pythonic code favors simplicity, readability, and leveraging the language’s strengths for more maintainable and expressive software.
The post explains that building a simple LLM agent, an app that maintains context, calls the model, and uses tools, is surprisingly easy and educational. It highlights that the real challenge is context engineering, such as managing what information to keep or summarize and how to integrate external tools, and encourages everyone to try building one to understand the technology deeply.
The django-allauth package had seven vulnerabilities, including four major ones that enabled account impersonation by exploiting mutable or non-unique identifiers from Okta and NetIQ, token refresh for deactivated users, and unverified email acceptance from Notion. These flaws could allow attackers to impersonate users and abuse tokens, but all have been fixed in version 65.13.0.
A high-level overview of how to use the module compression, new in Python 3.14.
The talk videos from DjangoCon US 2025 are now available.
Learn how we integrate Vue and Django for a bullet proof deployment.
The talk videos from PyBay 2025 are now available.
The plan for CPython's Free-Threaded JIT aims to achieve a 5% speedup by Python 3.15 and 10% by 3.16 through incremental optimizations. Improvements include better profiling and debugger support via LLVM, enhanced region selection and codegen, register allocation, reference counting optimizations, and more advanced JIT techniques—all contributing to a faster, more production-ready JIT in future Python releases.
CPython 3.15 introduces significant improvements in decompression speed: Zstandard (zstd) decompression is now up to 25–30% faster, and zlib decompression is 10–15% faster for data larger than 1 MiB. These gains result from optimizations in the handling of output buffers during decompression.
The "Dependency Cutout" workflow pattern provides a structured way to temporarily rely on a patched version of an open source dependency without permanently forking it. It involves forking the dependency, applying urgent fixes for your application, using this modified version in production, and then upstreaming the fix and unwinding back to the official release once your patch has been merged and released, minimizing long-term divergence and technical debt.
Interesting Projects, Tools, and Libraries
Open-source AI hackers for your apps.
Send a phone call from AI agent, in an API call. Or, directly call the bot from the configured phone number!
Harbor is a framework from the creators of Terminal-Bench for evaluating and optimizing agents and language models.
Communicate with an LLM provider using a single interface.
httpmorph is a drop-in replacement for Python's requests library that uses a custom C implementation with BoringSSL instead of Python's standard HTTP stack.
Realtime browser events for Django + PostgreSQL.
Automate your mobile devices with natural language commands - an LLM agnostic mobile Agent.
Visualize your FastAPI endpoints, and explore them interactively.
An open-source SQL-Native memory engine for AI.
Upcoming Events and Webinars
There will be following talks
The Zen of Claude Code
Functional Reproducibility
There will be following talks
How I built my own personal TripAdvisor in just half a day
Python 3.14 (or "π-thon" for real nerd pythonists)
From OpenAI to DeepSeek: New Scaling Laws for LLMs that can Reason
The unsatisfying issue of AI's environmental impact
There will be a talk, Anti-Patterns in Software Development.
There will be a talk, Generators and duck typing.
There will be a talk, Surviving the Snake Pit: How to Spot Threats, Read the Room, and Keep Your Job.
There will be following talks
Unlocking the Black Box: Demystifying ML Models with Shapley Values
Digital attribution model using LSTM
There will be following talks
From code to production: how to keep pipelines running smoothly at scale
Learnings from integrating AI within CyberBench: the good, the bad and the ugly
Our Other Newsletters |
Programmer Weekly - A free weekly newsletter for programmers.
Founder Weekly - A free weekly newsletter for entrepreneurs featuring best curated content, must read articles, how to guides, tips and tricks, resources, events and more.