PythonHub

How I built a sub-500ms latency voice agent from scratch The post walks through building a sub-500ms voice agent from scratch, with the main challenge being turn-taking: knowing exactly when the user is speaking versus listening so the agent can stop or respond instantly. It shows how the author simplified the system into a small streaming loop, then optimized latency with better model choice, geography, and a direct audio pipeline to make the... https://www.ntik.me/posts/voice-agent

tilelang Domain-specific language designed to streamline the development of high-performance GPU/CPU/Accelerators kernels. https://github.com/tile-ai/tilelang

Broccoli AI teammates for your engineering loop. Broccoli turns Linear tickets into shipped PRs - powered by Claude and Codex, running on your own Google Cloud. https://github.com/besimple-oss/broccoli

Python Hub Weekly Digest for 2026-05-03 https://pythonhub.dev/digest/2026-05-03/

Building a Python Library in 2026 So you want to build a Python library in 2026? Here's everything you need to know about the state of the art. https://stephenlf.dev/blog/python-library-in-2026/

Exploring Petabytes of the Night Sky — Jupyter Notebooks at NOIRLab’s Astro Data Lab Science Platform The post shows how NOIRLab’s Astro Data Lab uses Jupyter notebooks to let astronomers explore and analyze petabytes of sky data directly in the browser, without local setup. It also highlights the value of notebooks for making large-scale astronomy workflows more interactive, reproducible, and accessible to researchers and students. https://blog.jupyter.org/exploring-petabytes-of-the-night-sky-jupyter-notebooks-at-noirlabs-astro-data-lab-science-ae012dfd4723

ai-engineering-from-scratch From linear algebra to autonomous agent swarms. learn AI with AI, then ship the tools. https://github.com/rohitg00/ai-engineering-from-scratch

Alishahryar1 / free-claude-code Use claude-code for free in the terminal, VSCode extension or via discord like openclaw https://github.com/Alishahryar1/free-claude-code

What’s a low memory way to run a Python http endpoint? https://www.reddit.com/r/Python/comments/1slbp28/whats_a_low_memory_way_to_run_a_python_http/

PyWry: Cross-Platform Rendering Engine in Python https://deeleeramone.github.io/PyWry/

Dinobase Dinobase is an agent-first data platform that syncs 100+ sources like APIs, databases, files, and MCP servers into SQL-ready tables with automatic data annotation. https://github.com/DinobaseHQ/dinobase

Implementing MikroTik's Binary API Protocol in Python from Scratch A deep dive into implementing MikroTik's proprietary RouterOS binary API protocol in Python — variable-length encoding, sentence-based messaging, and programmatic network infrastructure control. Zero dependencies, 137 lines. https://www.joekarlsson.com/blog/implementing-mikrotik-binary-api-protocol-in-python/

Community consensus on when to use dataclasses vs non-OO types? https://www.reddit.com/r/Python/comments/1s7v9ug/community_consensus_on_when_to_use_dataclasses_vs/

uv or pip for python package management? https://www.reddit.com/r/learnpython/comments/1snqpnx/uv_or_pip_for_python_package_management/

Array API adoption: Performance wins across the ecosystem Adopting the Array API standard lets major Python libraries run the same code across backends like NumPy, PyTorch, CuPy, and JAX, unlocking dramatic speedups with minimal user changes. The broader impact is a more interoperable scientific Python ecosystem where GPU acceleration and new hardware become accessible without rewriting entire libraries. https://labs.quansight.org/blog/array-api-meta-blogpost

What Most Python Developers Miss About Generators Most Python developers view generators as a memory optimization, but their deeper value is controlling when computation happens and how data flows through a system. They enable lazy pipelines, backpressure handling, two-way communication, and patterns that extend naturally into async streaming architectures. https://www.youtube.com/watch?v=5VN-3rIUPZ8

Reimagine Python Notebooks in the AI Era Traditional notebooks are evolving as AI shifts more value from writing syntax to guiding workflows, reviewing outputs, and iterating through natural-language prompts. The emerging model emphasizes reactive cells and integrated LLM tooling that can turn linear notebooks into more interactive, dynamic applications. https://mljar.com/blog/reimagine-python-notebook-in-ai-era/

Best Python framework for industry-level desktop app? (PySide/PyQt/wxPython/Kivy/Web approacg) https://www.reddit.com/r/Python/comments/1s9oh61/best_python_framework_for_industrylevel_desktop/

lingbot-map A feed-forward 3D foundation model for reconstructing scenes from streaming data. https://github.com/Robbyant/lingbot-map

I built a dev blog! First deep dive: How Ruff and UV changed my mind about Python setups. https://www.reddit.com/r/Python/comments/1s7lna7/i_built_a_dev_blog_first_deep_dive_how_ruff_and/