- Python Weekly
- Posts
- Python Weekly (Issue 764 September 24 2026)
Python Weekly (Issue 764 September 24 2026)
Rahul Chaudhary
September 24, 2026
In partnership with
Welcome to issue 764 of Python Weekly. Let's get straight to the links this week.
The AI brief curated by Anthropic and ex-Google engineers
Thousands of AI papers, model releases, and product launches ship every week. You don't need all of them. You need the ten that matter.
TLDR AI is the free daily newsletter curated by Anthropic and ex-Google engineers. People who build AI for a living read everything, then send you what's actually worth your time: the models, research, and tools that will still matter next month.
No hot takes, no hype cycles, no 60-tweet threads. Clear summaries you can read between meetings.
Join 1.1M+ readers. Subscribe for free.
Articles, Tutorials and Talks
Tom Alard details how he built a highly competitive Ultimate Tic-Tac-Toe bot using a neural network trained on over 300 million self-play positions, a custom search algorithm, and SIMD-optimized C code. He also explains how he compressed the engine and neural network into a Python submission using UTF-16 encoding to bypass CodinGame’s 100,000-character limit, reaching second place on the leaderboard at the time of writing.
Timofei Ivankov explores how CinderX’s JIT compiler and Static Python can accelerate real-world Python services, comparing their internals and performance with CPython 3.14’s experimental JIT. Benchmarks show that combining Static Python with JIT compilation increased a CPU-bound endpoint’s throughput from 140 to 250 requests per second, while NumPy-heavy workloads saw no benefit and garbage-collection improvements varied by workload.
The author uses five exercises from two real LLM judges to expose evaluation pitfalls, including inconsistent results, biased test sets, misleading metrics, and pass/fail thresholds that become unreliable as test suites grow. He shows why trustworthy evaluations require representative data, clearly defined metrics, repeated testing, and preserved run artifacts, revealing flaws in his own evaluation tooling along the way.
Cloudflare has made Python Workers generally available, making Python a first-class language on the Workers platform. Developers can run frameworks such as FastAPI, Django, and Flask, use Cloudflare services like R2, D1, Durable Objects, Queues, Workflows, and Workers AI directly from Python, and avoid the JavaScript conversion glue previously required at runtime boundaries.
The post introduces Django Probe, a tool that scans Django projects and anonymously aggregates how framework APIs and patterns are actually used. The goal is to give Django maintainers better data for decisions around deprecations, new features, documentation, and community priorities.
This video compares three ways to model type-based variation in Python: subclasses, storing a type value such as an enum, and representing each variation as an object. Using a subscription system, it introduces the Type Object pattern and explains when each approach is the better fit.
This video compares three Python frameworks for building more capable AI agents that can navigate codebases, edit files, and run commands: CrewAI, the OpenAI Agents SDK, and LangGraph. It walks through building an agent with each framework and compares their approaches to orchestration, tools, workflows, and choosing the right framework for a project.
Nod (Notebooks-on-Demand) is a JupyterLab extension that lets you open an interactive notebook from anywhere in a running Python program with the current variables and execution context already available. Developers can inspect and modify code in the notebook, navigate the call stack, experiment with captured values, and send changes back to the original Python source file.
The article explains how vLLM is introducing hardware-agnostic layers so it can keep supporting diverse models and accelerators even as frontier models increasingly rely on hardware-specific “flat” implementations. The new path remains compatible with torch.compile and, in tests on NVIDIA H100s, delivered total token throughput within 3.4% of the native implementation across three recent models.
The article shows how to add an MCP server to a Python-based AWS cost scanner so AI agents can query computed totals, breakdowns, filters, and cleanup plans without doing arithmetic themselves. It also covers JSON-RPC over stdio, read-only tool design, end-to-end testing, rounding consistency, and integration with Claude Code.
Interesting Projects, Tools, and Libraries
tiny Jev-like family of decision models built on top of Qwen3.5 you can train and run on your own.
OpenRouter for agent tools.
A browser agent with a dynamic, indexed action space.
DeepTeam is a framework to red team LLMs and AI agents.
Continual learning infra for self-improving agents.
ASC is a super FAST Android decompiler front-end designed for Agents/Mobile Researchers.
TATS is an open-source Python tool for analyzing OAuth 2.0, OpenID Connect, and Microsoft Entra ID tokens, helping security researchers trace authentication flows, identify risky permissions, and visualize token lifecycles.
Ouroboros is an open-source AI coding framework that turns vague ideas into verified code through structured interviews, specification-driven execution, automated evaluations, and iterative improvement across multiple coding agents
The open-source agentic framework to build, orchestrate, and operate production AI agents.
Write Redis Lua scripts as real Python functions, not strings.
New Releases
PyPy 8.0.0 introduces its first Python 3.12 interpreter as a beta, alongside Python 2.7 and 3.11 releases, and raises the minimum glibc requirement for Linux binaries to 2.28. The release also advances compatibility with CPython’s limited C API and abi3 wheels, improves RPython code generation, and drops HPy as a default backend, though abi3 wheel installation support is not yet complete.
Upcoming Events and Webinars
There will be following talks
Building Agentic AI Systems
The physics behind (self-)hosting LLMs
Detecting and Mitigating Bias in Automated Mortgage Lending
There will be a talk, Satellite Data, Open Models, Geospatial AI, Forecasting, and AI Agents for Climate Action.
PyData Tallinn Meetup September 2026
There will be following talks
Building a Stable R&D process in a ML Team of a Frontier Physics Startup
Beyond the Quantum Hype: Enefit’s Journey into Quantum Computing
There will be following talks
How I AI: A Data Scientist's Perspective
The Data Scientist's Job in the Age of AI: What Changed, What Didn't
Trust in the Loop: System Design and Data Integrity in Regulated Environments
There will be following talks
From Changelog to Blog - AI Pipeline for Release Communication
Yet Another Year of Agents
There will be following talks
Mapping the PyData Community with Python and Web Scraping
10 years of Data Engineering with Python
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.