PythonHub

Agent Hooks: Deterministic Control for Agent Workflows The post is about agent hooks as a control layer that makes AI behavior more deterministic by enforcing rules at specific lifecycle points, instead of relying on prompts alone. It emphasizes using hooks for policy enforcement, validation, and observability so teams can block bad actions, add guardrails, and make agent workflows more production-safe. https://nader.substack.com/p/agent-hooks-deterministic-control

Watching for file changes on macOS This post details how to build a lightweight tool for monitoring file system activity on macOS without relying on third-party dependencies. It explains how to implement a Swift script that interfaces directly with Apple's native FSEvents API and forwards those live change notifications to a Python environment via stdout. https://alexwlchan.net/2026/watch-files-on-macos/

Python 3.15: features that didn't make the headlines https://blog.changs.co.uk/python-315-features-that-didnt-make-the-headlines.html

django-q2 - Background Tasks in Django (Celery alternative!) The video explores django-q2, a lightweight alternative to Celery for handling background jobs and asynchronous task processing in Django applications. It highlights how django-q2 simplifies setup and integration while providing built-in support for async workers, scheduled tasks, and Django Admin management features. https://www.youtube.com/watch?v=0hRDCrxfHug

geo-seo-claude GEO-first, SEO-supported. Optimize websites for AI-powered search engines. https://github.com/zubair-trabzada/geo-seo-claude

CodeBoarding Interactive architecture diagrams for codebases. https://github.com/CodeBoarding/CodeBoarding

django-completion Project-aware tab completion for Django's manage.py. https://github.com/soldatov-ss/django-completion

semble Fast and Accurate Code Search for Agents. Uses ~98% fewer tokens than grep+read. https://github.com/MinishLab/semble

nb-cli: A Command-Line Interface for AI Agents and Notebook Automation The article introduces nb-cli, a Rust-built command-line interface engineered to let developers and AI agents programmatically read, write, and execute Jupyter notebooks without needing a running server. It highlights an AI-optimized Markdown format that exposes structured token-efficient code and cell metadata, enabling language models and automated CI/CD pipelines to manipulate noteboo... https://blog.jupyter.org/nb-cli-a-command-line-interface-for-ai-agents-and-notebook-automation-996ad7edacd9

Numexpr: Fast numerical array expression evaluator for Python, NumPy, Pandas https://github.com/pydata/numexpr

Python Hub Weekly Digest for 2026-05-24 https://pythonhub.dev/digest/2026-05-24/

In-Kernel Broadcast Optimization: Co-Designing Kernels for RecSys Inference In-Kernel Broadcast Optimization (IKBO) eliminates redundant user-embedding replication by fusing broadcast logic directly into interaction kernels, significantly reducing memory bandwidth and compute waste. This co-design approach delivers up to a two-thirds reduction in latency across Meta's recommendation stack, optimized for high-performance hardware like NVIDIA H100 and Meta’s MTIA. https://pytorch.org/blog/in-kernel-broadcast-optimization-co-designing-kernels-for-recsys-inference/

OpenSwarm Claude code for everything except coding. https://github.com/VRSEN/OpenSwarm

aidlc-workflows AI-Driven Life Cycle (AI-DLC) adaptive workflow steering rules for AI coding agents. https://github.com/awslabs/aidlc-workflows

DRY Often Makes Your Code Worse A DRY refactor can make code worse when it removes duplication by hiding different behaviors behind flags and shared abstractions. Better refactoring focuses on separating responsibilities and making behavior clear, not just making the code look smaller. https://www.youtube.com/watch?v=GmlZBdKhl9Y

Labb - UI components for Django This video looks at labb - a UI component library for Django apps that builds on top of TailwindCSS, DaisyUI, django-cotton and Alpine.js. This library has over 70 components that can be dropped into Django projects, and makes it easy to add custom themes, for example using DaisyUI Theme Generator. https://www.youtube.com/watch?v=ZZd7cvbJ-1w

The Simplest MCP Example Possible in Python The post shows the simplest Python example of MCP (Model Context Protocol) for connecting a local LLM like Llama 3.2 to tools. It uses fastmcp and ollama packages with two scripts: mcpserver.py for time/date tools and ollamaclient.py to run the chat loop. https://inventwithpython.com/blog/basic-mcp-python-example.html

Introducing Retrace: Record Production Python. Debug It Backwards. Retrace introduces deterministic record-replay debugging for Python, allowing production executions to be recorded with very low overhead and replayed locally in VS Code with reverse debugging support. Instead of relying on logs or reproducing failures, it records the boundary between application code and nondeterministic systems like APIs, databases, threads, and file I/O, enabling deve... https://retracesoftware.com/blog/introducing-retrace/

Reverting the incremental GC in Python 3.14 and 3.15 https://discuss.python.org/t/reverting-the-incremental-gc-in-python-3-14-and-3-15/107014

Family Orienting Python Frozenset Dependent Type Theory The post models dependent type theory in plain Python by treating types as finite “families,” or dictionaries from contexts to sets, rather than as standalone bare values. The key idea is that full judgments like Γ |- A should be interpreted as mappings, which makes contexts, weakening, terms, Sigma/Pi types, and identity types easier to represent uniformly. https://www.philipzucker.com/dtt_python_family/