Python Hub Weekly Digest for 2026-09-13
This week in Python land, we’re serving up a menu of delights that’ll tickle your code-sense! Want to astound your interviewer with those tech company interview gems? We’ve got an ever-fresh collection that’s bound to impress—if only impressing bots counted for points! Meanwhile, don’t miss out on Linecast, a terminal app that’s like the Old Farmer's Almanac got a degree in meteorology. And for those who prefer to keep their finances as private as their chocolate stash, check out Securo, the open-source personal finance manager. May your week be bug-free and your commits be ever in your favor!
💖 Most Popular
Tech-OA-Interview-Questions
Daily updated list of Tech Company OAs and Interview Problems. Save your time from finding them all over the internet.
linecast
Weather, tides, the sun, the moon, and maps, in your terminal. The Old Farmer's Almanac meets Minitel.
Securo
Open-source personal finance manager. Self-hosted, privacy-first.
Python Hub Weekly Digest for 2026-09-06
Making a Python interpreter in 1024 bytes
📖 Articles
Training and Finetuning Multi-Vector Embedding Models with Sentence Transformers
Sentence Transformers 6.0 introduces MultiVectorEncoder for training and fine-tuning ColBERT-style retrieval models in Python. The tutorial walks through the full training stack and shows a domain-tuned medical model outperforming over 50 retrieval model configurations on the author’s evaluation.
Unit testing with wrapture
Wrapture takes a different approach to Python unit testing by wrapping real code instead of replacing it with mocks, preserving real arguments, return values, call structure, and signatures. It makes it easier to test internal calls, inject failures, verify call order, and assert what did or did not happen without losing the behavior of the code under test.
When to use NotImplemented
When should you return NotImplemented from a dunder method? Why not return False or raise an exception instead?
KV, Prefix, Prompt and Semantic Caching in LLMs, clearly explained
The post explains the four caching layers in LLM stacks: KV cache, prefix caching, prompt caching, and semantic caching, covering what each stores, their trade-offs, interactions, and the most common issues that prevent reuse. It includes first-principles explanations, code demos (with transformers and vLLM-style logic), production pitfalls, and practical takeaways for reducing recomputa...
Fine-Tuning SOTA Object Detection Models on Real-World Datasets
Learn how to use these models, how to fine-tune them on diverse, specialized datasets that look nothing like their training data, and how to evaluate the results – all within PyCharm.
index-tts / index-tts
An Industrial-Level Controllable and Efficient Zero-Shot Text-To-Speech System
Tracing np.add, all the way down
The author traces a simple np.add(a, b) call through NumPy’s internals, from Python argument handling and dtype dispatch to iteration strategy and the final SIMD-optimized C loop. Along the way, the post explores ufunc overrides, type promotion, dispatch caching, memory layout, GIL release, and CPU-specific optimizations that determine how NumPy actually performs an addition.
Nifty Django Feature: Use Index for Custom Migration Operations
Part of the Nifty Django features series: You can hack Django's Index class to define custom migration operations on your model.
K-Dense-AI / scientific-agent-skills
Turn any AI agent into an AI Scientist. The #1 Agent Skills library for science, used by 190,000+ scientists worldwide. 165 ready-to-use validated skills plus 100+ scientific databases covering biology, chemistry, medicine, and drug discovery. Compatible with Cursor, Claude Code, Codex, Pi, Antigravity, and the open Agent Skills standard.
⚙️ Projects
FlightScnr_Pi
Desktop flight and marine radar: a real-time aircraft and marine vessel tracker powered by a Raspberry Pi and 4" round screen.
experiential
An open source model gateway that provides one control plane across closed, open-source, local, and custom models.
Neocarta
Library built for generating semantic layer graphs for query routing, query generation and data discovery.
OpenExecutive
AI-powered virtual executive team, a single coherent executive persona backed by 8 specialist Claude agents (FastAPI + Next.js).
microduck_rl
RL training environments for Microduck (mjlab)
AutoSaddler
Automatic Harness Optimization with Durable Updates from Agent Execution Traces.
🎬 Videos
Building AI Agents in Pure Python - Beginner Course
Build a fully functioning AI agent from scratch in pure Python, without frameworks, third-party tools, or vibe coding. The course focuses on the core mechanics behind agents so you understand how they work under the hood, not just how to prompt existing tools.
You Think This Is Good OOP… It’s Not
Six common OOP practices can make code harder to maintain, including overusing inheritance, building unnecessary class hierarchies, breaking the Liskov Substitution Principle, and abstracting too early. The video explains why these patterns backfire and shows simpler alternatives that keep object-oriented code more flexible, understandable, and maintainable.
Django API Overview - DRF, django-ninja, django-bolt and django-modern-rest
In this video, we'll look at Django API options, exploring the best modern packages for API-driven development in Django. We will take a look at Django REST Framework, django-ninja, django-bolt, and django-modern-rest - looking at the pros/cons of each.
Project by Ruslan Keba. Since 2012. Powered by Python. Made in 🇺🇦Ukraine.