earthtojake / text-to-cad

A library of agent skills for CAD, CAE and CAM

https://github.com/earthtojake/text-to-cad

A quick look at zero-knowledge proofs

In this post, the authors break down non-cryptocurrency zero-knowledge proofs (ZKPs) using graph theory and Python. By implementing Protocol 4 from Goldreich, Micali, and Wigderson, they show how a prover uses randomized color permutations, nonces, and cryptographic hashes to iteratively prove they hold a valid 3-coloring for a graph without revealing the actual solution to the verifier.

https://bernsteinbear.com/blog/zkp/

From Routing Checks to Trajectory Testing: Evaluating an Agentic Chatbot

Traditional chatbot testing that checks only final responses misses many agent failures, so this article builds an evaluation framework that progresses from routing checks and task completion to full trajectory testing of tool use, conversations, and intermediate actions. It also shows how to combine LLM-as-a-judge evals, multi-turn testing, and telemetry to debug agent behavior and vali...

https://edcrewe.blogspot.com/2026/08/from-routing-checks-to-trajectory.html

interlock

Circuit breaker for Python: sync + async in one class, sliding-window rate, slow-call detection, Polly-style resilience pipeline, type-safe API.

https://github.com/bagowix/interlock

Two lines of Python that segfault the interpreter

Two lines of Python could crash CPython 3.14 through 3.16 because SETADD assumed it was always operating on a real set, while PEP 749 madeconditionalannotationsrebindable from Python code. The post explains how that assumption broke, why it caused a type-confusion crash, and why stable releases and the development branch needed different fixes.

https://deadlovelll.github.io/2026-08-10-conditional-annotations-set-add-crash/

Composite: The Pattern Behind Menus, File Systems and Games

The Composite pattern lets you treat individual objects and entire object hierarchies through the same interface, making tree-like structures much easier to manage. Using a simple Python game engine, the video demonstrates how this pattern naturally applies to game scenes, UI frameworks, file systems, menus, and similar hierarchical designs.

https://www.youtube.com/watch?v=ss6je4-nDx8

Numba in the Browser: Unlocking a New Scientific Python Stack in JupyterLite

Numba now runs entirely in the browser through JupyterLite, compiling Python functions to WebAssembly with a new LLVM-based execution engine and delivering substantial performance gains without a server. The work also unlocks browser-native support for the broader Numba ecosystem, including PyMC, PyTensor, and other scientific computing libraries.

https://notebook.link/blog/numba-in-the-browser/

python-game-server

A lightweight server and framework for turn-based multiplayer games.

https://github.com/feberts/python-game-server

key-amnesia

Let your AI agent use your passwords and API keys - without ever letting it see them.

https://github.com/fujitoid/key-amnesia

Prototype on a laptop, scale to 16 billion rows: one Polars query

The post shows how to prototype a data pipeline locally on 97 million Polymarket orderbook rows, then run the same LazyFrame queries on 16 billion rows using distributed execution. The pipeline pre-aggregates the raw data into small Parquet artifacts in S3 that power a responsive Plotly Dash dashboard without scanning the full dataset on each request.

https://pola.rs/posts/market-data-to-plotly-enterprise-dashboard/

jsonfold: Making Pretty-Printed JSON Compact and Readable in Python

jsonfold is a streaming post-filter that compacts pretty-printed JSON by folding small arrays and objects onto single lines while preserving existing serializers and custom encoders. It processes JSON incrementally without reparsing the document, keeping additional memory usage bounded while producing more compact, human-readable output for large JSON files.

https://medium.com/@yair.lenga/a-streaming-json-formatter-that-works-with-existing-serializers-eced220da37d

centaur

Centaur is frontier, agentic infrastructure that you own. Centaur is like Claude Tag, but open source and on steroids.

https://github.com/paradigmxyz/centaur

Comfy-Org / ComfyUI

The most powerful and modular diffusion model GUI, api and backend with a graph/nodes interface.

https://github.com/Comfy-Org/ComfyUI

pyhctsa

The most comprehensive time-series feature extraction package in Python.

https://github.com/DynamicsAndNeuralSystems/pyhctsa

django-waffle - the best Feature Flipper for Django!

This is an introduction to django-waffle, a production-ready Django package for controlling feature rollouts through flags, switches, and sampling. It enables gradual releases, A/B testing, and quick kill switches for features that need to be disabled without redeploying

https://www.youtube.com/watch?v=K-Y9XHEZwU8

Tencent opensource AI Red Teaming platform

A full-stack AI Red Teaming platform securing AI ecosystems via Agent Scan, Skills Scan, MCP scan, AI Infra scan and LLM jailbreak evaluation.

https://github.com/tencent/AI-Infra-Guard

Lians

Evidence-backed proof of done for Claude Code, Codex, Cursor, and other AI coding agents. Run real checks, bind results to current Git state, and know what is ready for human review.

https://github.com/Lians-ai/Lians

The Polars vs SQL differences nobody is talking about

This post explores important differences between Polars and SQL that can cause unexpected results when moving between the two, including row ordering, column independence, literals, null handling, and broadcasting. It explains how understanding these different mental models can help developers write safer Polars code and more reliably translate data workflows between Polars and SQL.

https://labs.quansight.org/blog/polars-vs-sql-differences