open-terminal

A lightweight, self-hosted terminal that gives AI agents and automation tools a dedicated environment to run commands, manage files, and execute code — all through a simple API.

https://github.com/open-webui/open-terminal

Starlette 1.0

After nearly eight years since its creation, Starlette has reached its first stable release. Today, it's downloaded almost 10 million times a day, serves as the foundation for FastAPI, and has inspired many other frameworks. In the age of AI, Starlette continues to play an important role as a dependency of the Python MCP SDK.

https://marcelotryle.com/blog/2026/03/22/starlette-10-is-here/

ProperDocs

ProperDocs is a static site generator intended for project documentation. Source files are written in Markdown and converted to static HTML during the build process.

https://github.com/ProperDocs/properdocs

Pydantic AI intro - Agents and Instructions!

In this video, we'll take a look at Agents in Pydantic AI, and will create a simple Agent that can be run synchronously and asynchronously. We'll explore how to inspect agent outputs, and how to amend outputs using instructions that are passed to the LLM.

https://www.youtube.com/watch?v=XyGHYG7QNK0

Terrapod

Open-source Terraform Enterprise replacement.

https://github.com/mattrobinsonsre/terrapod

Lessons from Pyre that Shaped Pyrefly

High-performance type checking at Meta required a performance-first architecture and tight integration with developer workflows, enabling fast, incremental analysis at massive scale. The key lesson is that large Python codebases adopt typing successfully through gradual, low-friction tooling that prioritizes developer ergonomics and fast feedback over strict correctness.

https://pyrefly.org/blog/lessons-from-pyre/

MiniStack

LocalStack is no longer free. MiniStack is a fully open-source, zero-cost drop-in replacement. Single port · No account · No license key · No telemetry · Just AWS APIs, locally.

https://github.com/Nahuel990/ministack

Rewriting a 20-year-old Python library

The article covers a full rewrite of the Akismet Python client to add async support, modern HTTP handling, and a richer response model while preserving usability. It emphasizes API ergonomics, testing support, and maintainability, while honoring the original author and evolving the library for modern Python.

https://www.b-list.org/weblog/2026/mar/23/20-year-library/

voicetag

Speaker identification powered by pyannote and resemblyzer.

https://github.com/Gr122lyBr/voicetag

slamd

A 3D visualization library for Python. pip install, write a few lines, and you have a GPU-accelerated interactive 3D viewer.

https://github.com/Robertleoj/slamd

Deploying AI Models with Hugging Face – Hands-On Course

This tutorial is a comprehensive, end-to-end guide to the Hugging Face ecosystem, showing how modern AI moves from research ideas to real, deployed systems. Rather than focusing on a single model or task, the course presents Hugging Face as the operating system of modern AI—connecting models, datasets, libraries, demos, and deployment into one coherent, practical workflow.

https://www.youtube.com/watch?v=R8h_gpSpEVU

Python Type Checker Comparison: Typing Spec Conformance

When you write typed Python, you expect your type checker to follow the rules of the language. But how closely do today's type checkers actually follow the Python typing specification? In this post, we look at what typing spec conformance means, how different type checkers compare, and what the conformance numbers don't tell you.

https://pyrefly.org/blog/typing-conformance-comparison/

Using Claude to fix PyPy3.11 test failures securely

This post describes using Claude to assist in fixing PyPy 3.11 test failures, with all generated changes run in a sandbox and verified locally. It highlights a practical workflow where AI suggests patches but humans validate results, enabling faster debugging without sacrificing safety.

https://pypy.org/posts/2026/03/using-claude-to-fix-pypy311-test-failures-securely.html

How we optimized Dash's relevance judge with DSPy

Dropbox used DSPy to turn prompt engineering for our relevance judge into a measurable, automated optimization loop, improving task performance, cost, and how reliably it works in production.

https://dropbox.tech/machine-learning/optimizing-dropbox-dash-relevance-judge-with-dspy

Reinventing Python's AsyncIO

The post explores a redesign of Python’s async runtime, arguing that the current async/await and event-loop model adds unnecessary complexity, and proposing a simpler runtime where concurrency is handled automatically without explicit async syntax.The author experiments with a new runtime approach that can run async workloads 2–3.5× faster than traditional asyncio, suggesting Python’s co...

https://blog.baro.dev/p/reinventing-pythons-asyncio