Recursive Language Models

Recursive Language Models (RLMs) let language models recursively call themselves within an environment, like a Python REPL, to handle extremely long contexts without performance drop (context rot). They dynamically break down queries into smaller parts, delivering strong, cost-efficient results on big benchmarks and enabling scalable, interpretable reasoning beyond fixed context limits.

https://alexzhang13.github.io/blog/2025/rlm/

Introducing PyTorch Monarch

PyTorch Monarch is a distributed programming framework designed to simplify scaling AI workflows by enabling a single-controller model that orchestrates distributed resources like a single machine. It provides actor-based programming with scalable messaging, fault tolerance, and distributed tensor support, allowing seamless development, debugging, and efficient handling of large-scale tr...

https://pytorch.org/blog/introducing-pytorch-monarch/

Katakate

Self-hosted secure VM sandboxes for AI compute at scale.

https://github.com/Katakate/k7

caniscrape

Know before you scrape. Analyze any website's anti-bot protections in seconds.

https://github.com/ZA1815/caniscrape

Create Your Own Bash Computer Use Agent with NVIDIA Nemotron in One Hour

A tutorial on building a computer use AI agent capable of executing multi-step tasks in a Bash shell, powered by the NVIDIA Nemotron Large Language Model. It covers creating the agent's brain, the Bash interface for safe command execution, and the agent loop, demonstrating how to build and deploy an autonomous assistant within an hour.

https://www.youtube.com/watch?v=F7f-eFou2-o

Dexter

An autonomous agent for deep financial research

https://github.com/virattt/dexter

Blinter

Blinter is a linter for Windows batch files. It provides comprehensive static analysis to identify syntax errors, security vulnerabilities, performance issues and style problems.

https://github.com/tboy1337/Blinter

django-bolt

Rust-powered API framework for Django achieving 60k+ RPS. Uses Actix Web for HTTP, PyO3 for Python bridging, msgspec for serialization. Decorator-based routing with built-in auth and middleware.

https://github.com/FarhanAliRaza/django-bolt

Wrapping immutable objects

Graham Dumpleton explores how Python’s wrapt library must correctly handle immutable types when wrapping objects, highlighting hidden issues with in-place operators like += and @= when applied to proxies of immutables. He outlines a refinediaddstrategy for ObjectProxy that preserves expected behaviour in version 2.0.0 of wrapt.

https://grahamdumpleton.me/posts/2025/10/wrapping-immutable-types/

typeagent-py

This is an in-progress project aiming at a Pythonic translation of TypeAgent KnowPro and a few related packages from TypeScript to Python.

https://github.com/microsoft/typeagent-py

burla

Run any Python function on 1000 computers in 1 second.

https://github.com/Burla-Cloud/burla

Brewing with SQLAlchemy

This series is a comprehensive workshop designed to teach SQLAlchemy from fundamental concepts through advanced data modeling and complex query crafting. By the end, it aims to equip learners with practical skills to confidently handle data challenges and build scalable, powerful applications using SQLAlchemy’s ORM and session features.

https://fullstack.rocks/article/sqlalchemy/brewing_with_sqlalchemy

wove

Beautiful Python async. It is for running high latency async tasks like web requests and database queries concurrently in the same way as asyncio, but with a drastically improved user experience.

https://github.com/curvedinf/wove

How DeepSeek OCR Quietly Solved a Billion-Dollar Problem in AI Scaling

This article cover the architecture and how to guide for DeepSeek OCR- this story has gone viral on this topic.

https://medium.com/data-and-beyond/how-deepseek-ocr-quietly-solved-a-billion-dollar-problem-in-ai-scaling-7b4502613af9?sk=d43719a9caf7dfe938f9af1eac86056b

Monte Carlo modeling in Python with probabilit

The article introduces probabilit, a Python package designed for Monte Carlo modeling that simplifies uncertainty calculations by allowing users to model equations with probability distributions and sample from them. It demonstrates key features like Latin Hypercube Sampling for efficient sampling and inducing correlations between variables, highlighting its suitability for prototyping u...

https://tommyodland.com/articles/2025/monte-carlo-modeling-in-python-with-probabilit/

skylos

Yet another static analysis tool for Python codebases written in Python that detects dead code + common security flaws created by ai. Faster and better than the rest :) also, who let the dawgs out?

https://github.com/duriantaco/skylos