Introducing Flex: Let the Model Write the Code

DSPy’s new Flex module allows optimizers like GEPA to improve programs by rewriting both their instructions and underlying Python code. This lets the optimizer route easy cases to fast deterministic code while reserving model calls for ambiguity, significantly reducing cost and latency while improving accuracy.

https://www.cmpnd.ai/blog/let-the-model-write-the-code.html

Agent Memory Guard

Agent Memory Guard is an OWASP Incubator Project that prevents AI agents from being weaponized through their own memory. It implements MITRE ATLAS mitigation AML.M0031 (Memory Hardening) to defend against context poisoning attacks (AML.T0080).

https://github.com/OWASP/www-project-agent-memory-guard

Model Genome: Fingerprinting Whether an LLM Was Trained From Scratch or Derived

Outsiders can assess whether a foundation model was truly built from scratch by analyzing architecture configurations, tokenizer overlap, and weight embeddings using a reproducible fingerprinting pipeline. While architecture and tokenizer artifacts provide the strongest evidence, weight analysis has limitations and cannot cleanly distinguish continued pretraining from training from scratch.

https://huggingface.co/blog/mayafree/model-dna

semantica

Graph-Native Infrastructure for Context and Accountable AI Systems.

https://github.com/semantica-agi/semantica

Python: introducing emojet, a fast emoji lookup library

emojet is an emoji library for Python: it converts between emoji and their names, in both directions, plus the searching and lookup functions that go with that. It covers the core API of the emoji package, a library that been available for this job since 2014, using the same names and the same data. The difference is that emojet does the work in Rust, running 3.5x faster for conversion, ...

https://adamj.eu/tech/2026/08/12/python-introducing-emojet/

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