Tag
A paper showing that running statistics over raw LLM judge scores inflates false positives, introducing PPI-based calibrated hypothesis tests (including the first PPI Wilcoxon/Mann-Whitney corrections) and the open-source Python package evalstats, which automatically selects statistically sound methods for small-sample AI evaluations.
Livenerf is a deterministic benchmark tool designed to detect if AI models degrade after launch, currently running a 30-day study on Claude Opus 5.5 to measure performance drift.
This article explains the rounding errors in decimal calculations with floating-point numbers in programming, using Python examples to illustrate the patterns that emerge from IEEE double-precision floats.
This article introduces 'Agent + Skill Store', a local AI agent tool that runs on your machine with a headless browser, sandboxed shell execution, and a plugin system for adding skills without cloud dependencies.
TimeLord is a Python tool that constructs seeds for CPython's PRNG to produce specified sequences, demonstrating how improbable outcomes can be engineered using deterministic generators.
Jev is a new AI model optimized for fast and cheap decision-making in classification tasks, demonstrated with a hotel review classifier project using Apify and GPT-5-mini.
The Python Coding Assistant is an AI agent that generates production-grade Python code with type hints, tests, and framework awareness, providing targeted fixes for libraries like FastAPI and pandas while remembering the user's stack across sessions.
A developer created a library called apple-llm to make Apple's built-in local LLM on macOS easily usable from Node and Python, focusing on privacy and specific tasks like structured data extraction and summarization.
The article demonstrates how to solve a physical corn puzzle using Google's OR-Tools CP-SAT constraint satisfaction solver, highlighting its efficiency over naive backtracking methods.
Hyperresearch is an open-source Python tool that acts as a deep research harness for Claude Code and OpenAI Codex, using agents to collect, search, and synthesize web research into a persistent, searchable wiki with adversarial auditing.
This article explains how to use the logprobs parameter in LLMs to implement Jev-style encapsulation for supporting structured queries, and extend to visual models via the attachments field for image processing, offering a flexible and customizable computer vision approach.
A terminal port management tool written in pure Python that scans and organizes listening ports, processes, projects, and containers into a unified view, showing service origins and whether they are externally accessible.
The article describes a Python wrapper for LLMs that uses token probabilities for classification tasks, extended to vision models with an example for real-time webcam frame analysis.
The post announces the release of lm15 1.0 on PyPI, a lightweight, zero-dependency Python SDK for provider-agnostic LLM calls, which is open-source and MIT licensed.
Citation-Network v1.2.0 is an open-source app that builds interactive citation networks from Zotero libraries, with integration for Obsidian notes and a GUI for exploration.
pylazaro is an open-source Python package for automatic extraction of unassimilated lexical borrowings (mostly anglicisms) from Spanish text, offering a unified interface to multiple sequence labeling models and outperforming general-purpose LLMs on this task.
A TUI text adventure tool that guides users through the new features in Python 3.15.
DeskPilot is an open-source Python-based desktop client for local LLMs, featuring MCP integration, tool sandboxing, and a security-first agentic architecture.
The article announces the release of llm-anthropic version 0.29, a Python tool for interacting with Anthropic's API, published by Simon Willison on September 22, 2026.
Kent Dodds experimented with running Python on Kody using Cloudflare's GA for Python in Workers, but decided against implementing it due to slow startup times.