bigsnarfdude

name	description	license
skill-creator	Guide for creating effective skills. This skill should be used when users want to create a new skill (or update an existing skill) that extends Claude's capabilities with specialized knowledge, workflows, or tool integrations.	Complete terms in LICENSE.txt

Skill Creator

This skill provides guidance for creating effective skills.

bigsnarfdude

🐍 Algesnake - 39 Working Examples

  basic_example.py (6 examples)

  Foundation concepts using abstract algebra:
  - Example 1: Max Monoid - Finding maximum values
  - Example 2: Integer Addition Group - Addition with subtraction
  - Example 3: Ring Operations - Distributivity (a × (b + c))
  - Example 4: Complex Ring Expressions - Combined operations
  - Example 5: MonoidWrapper - Quick prototyping without classes

bigsnarfdude

# LLM brain rot reveals persistent cognitive decay from junk data

**LLMs exposed to viral web content suffer lasting cognitive decline that resists standard fixes.** A groundbreaking October 2025 paper demonstrates that continual training on engagement-driven junk data causes thought-skipping, safety failures, and personality distortion—damage that instruction tuning and clean data retraining only partially reverse. This "brain rot" persists because junk data fundamentally rewires model representations at the weight level, creating shortcuts that become deeply entrenched. The finding has critical implications: data quality is a training-time safety issue, not just a performance optimization, and deployed models may be silently degrading as they ingest low-quality web content.

The research connects to broader phenomena including model collapse, where AI systems recursively trained on synthetic data lose information about rare events, and sleeper agent backdoors that resist removal through safety training. To

bigsnarfdude

No legal mandate for independent monitoring of monitoring systems
No requirement that AI systems declare their dependencies on other AI systems
No standard for cross-system risk assessment when outputs feed into other systems
No accountability structure when cascade failures happen across multiple organizations
No legal requirement that someone be actively responsible for detecting failures in real-time

bigsnarfdude

---
name: docx
description: "Comprehensive document creation, editing, and analysis with support for tracked changes, comments, formatting preservation, and text extraction. When Claude needs to work with professional documents (.docx files) for: (1) Creating new documents, (2) Modifying or editing content, (3) Working with tracked changes, (4) Adding comments, or any other document tasks"
license: Proprietary. LICENSE.txt has complete terms
---

# DOCX creation, editing, and analysis

## Overview

bigsnarfdude

#!/usr/bin/env python3
# -*- coding: utf-8 -*-
"""
GPT-OSS Reinforcement Learning for 2048 Game - FIXED VERSION

Key fix: extract_function now correctly extracts generated code instead of the example
"""

# CRITICAL: Import unsloth FIRST before any other ML libraries
from unsloth import FastLanguageModel

bigsnarfdude

# Vision-Language Models transform robot control systems

Vision-Language-Action (VLA) models have revolutionized robotic manipulation since 2023, achieving unprecedented generalization by training unified models that process images and natural language to generate robot actions at scale. The field has rapidly progressed from Google DeepMind's RT-2 demonstrating 63% improvement on novel object manipulation to open-source models like OpenVLA outperforming proprietary systems with **7x fewer parameters**, while Physical Intelligence's π0 enables **50Hz real-time control** for complex tasks like laundry folding. This paradigm shift from specialized controllers to general-purpose policies trained on millions of demonstrations across 22+ robot platforms represents the most significant advance in robotics autonomy, with models now deployed in industrial settings from BMW factories to commercial kitchens.

The transformation stems from three breakthrough insights: robot actions can be represented as language tokens 

bigsnarfdude

#!/usr/bin/env python3
"""
IPA Translator using Ollama with Gemma3
Converts English text to International Phonetic Alphabet notation
"""

import requests
import json
import sys
import re

bigsnarfdude

# Install all required dependencies with non-interactive setup
print("Setting up non-interactive environment...")
!echo 'debconf debconf/frontend select Noninteractive' | sudo debconf-set-selections
!echo "keyboard-configuration keyboard-configuration/layoutcode string us" | sudo debconf-set-selections
!echo "keyboard-configuration keyboard-configuration/variantcode string" | sudo debconf-set-selections

print("Installing system dependencies...")
!curl -fsSL https://ollama.ai/install.sh | sudo sh
!sudo apt-get update -y
!sudo DEBIAN_FRONTEND=noninteractive apt-get install -y cuda-drivers ocl-icd-opencl-dev nvidia-cuda-toolkit

bigsnarfdude

#!/usr/bin/env python3
"""
MCP Server with Compound Interest Calculator Tool - Python Version

This is the Python equivalent of the JavaScript MCP server shown in the image.
It provides a compound interest calculation tool using the FastMCP framework.

To run this server:
1. Install dependencies: pip install "mcp[cli]"
2. Run directly: python compound_interest_server.py