- Python Weekly
- Posts
- Python Weekly (Issue 765 October 1 2026)
Python Weekly (Issue 765 October 1 2026)
Rahul Chaudhary
October 01, 2026
Welcome to issue 765 of Python Weekly. Let's get straight to the links this week.
If you want to sound smarter without doing more…
Read Morning Brew. 5-minute daily email. Day's most important news. Explained in simplest terms. Really, it’s a news experience you’ll actually enjoy.
At this point, it's not about trying Morning Brew's daily newsletter—It's about missing out.
Click below to join 4,000,000+ daily readers.
News
Anthropic’s MCP Python SDK had an OAuth vulnerability that could let malicious MCP servers steal credentials and potentially take over user accounts. The issue affected several SDK versions and authentication providers and was fixed in mcp 2.2.0 and 1.30.0.
Articles, Tutorials and Talks
Build AI systems from first principles with 523 free, open-source lessons spanning math, models, tokenizers, agents, evaluation, and production.
Philip Zucker shows how a Python union-find data structure can generate Lean proof certificates, turning union and path-compression operations into machine-checkable equality proofs. He uses the experiment to explore the broader idea of treating proofs as traces of search, with applications to e-graphs and theorem-proving systems.
Quail is an open-source AI-SQL engine that jointly optimizes query planning and LLM inference to reduce redundant model work and keep GPUs busy when running large numbers of AI-powered database operations. In benchmarks across 29 queries, it was 1.84x faster on average than tuned vLLM baselines and up to 14.04x faster on a workload with substantial KV-cache reuse.
This course demonstrates how to build an algorithmic paper trading system using Python and Django to calculate momentum scores and execute trades. The project integrates Massive for market data, SnapTrade for portfolio management, and Alpaca for risk-free brokerage simulation
The post shows how to serve a standards-compliant security.txt file in Django so security researchers know how to report vulnerabilities. It also adds tests and a Django system check to validate the file and warn before its required Expires field becomes stale.
Pyrefly is Meta’s open-source static type checker for Python, designed to catch type-related bugs before code runs instead of waiting for runtime failures. The video explains why Python and type hints alone can miss these errors, and shows how Pyrefly detects them early in real-world codebases.
A concise history of text classification, from bag-of-words and neural networks to transformers and modern LLM-based approaches. It also introduces Jev and examines where specialized classifiers can still outperform or complement general-purpose language models.
The video shows how to write custom FastAPI middleware to inspect, modify, or block requests before they reach route handlers using the call_next pattern. It also briefly covers built-in middleware for HTTPS redirects, trusted hosts, and GZip compression.
This video serves as a comprehensive crash course on LangSmith, the LangChain ecosystem platform used for tracing, debugging, evaluating, and monitoring AI agents. It covers essential LLMOps workflows including tracing agent runs, performing evaluations with custom datasets, creating alerts, and managing prompts.
Interesting Projects, Tools, and Libraries
Run open decision models locally: pull and serve Laya, decider, NLI and GLiClass behind a TypeSafe-compatible API. Ollama for decision models.
Which of your dependencies block a Django upgrade, and in which order to upgrade them.
Taste-Bench measures the taste of an LLM agent: its ability to choose the better direction at a real decision fork in a long-horizon task.
Build an agent harness and control it end-to-end. Open-source SDK for production AI agents in Python & TypeScript - any model, any cloud.
Fine-tunes of Qwen3.5 and Gemma 4 for zero-shot classification.
A long-running, deterministic-as-possible benchmark for detecting whether a frontier model gets quietly worse after launch.
The open-source workspace where humans and AI agents work as one team.
Fast runtime type enforcement for Python 3.11+ type annotations. Zero dependencies and uncompromising performance.
Laya is a fast multilingual decision engine that produces typed classifications, scores, and probabilities in a single forward pass.
A smarter, self-hosted AI assistant - multi-user, multi-agent.
Upcoming Events and Webinars
There will be following talks
From PoC to Production: Building Scalable Agentic Applications
Listening to black holes with missing data
Making Python Faster
There will be following talks
Data problems in the greenhouse
Robot Whispering on the Construction Site
There will be a talk, Accuracy Isn’t Everything: Good Models are Lurking Everywhere.
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.