Advent of Claude 2025

A collection of solutions for the Advent of Code 2025 programming challenges, solved using pure vibe coding with Claude Code.

What is Pure Vibe Coding?

This repository is an experiment in AI-assisted development where all code is written by Claude Code with zero manual edits. Every solution, test, and documentation file is generated through natural language interaction with Claude Code, demonstrating the power of AI-powered development workflows.

Project Structure

AdventOfClaude-2025/
├── README.md           # This file
├── CLAUDE.md           # Instructions for Claude Code instances
├── mise.toml           # Tool versions and task definitions
└── days/               # All daily solutions
    ├── 01/             # Day 1 puzzle solution
    │   ├── INSTRUCTIONS.md    # Puzzle description
    │   ├── part_1.py          # Part 1 solution
    │   ├── part_2.py          # Part 2 solution
    │   ├── test.py            # Test suite
    │   ├── input.txt          # Puzzle input (user-specific)
    │   └── README.md          # Solution documentation
    ├── 02/             # Day 2 puzzle solution
    │   └── ...
    └── ...

Each day's directory contains:

• INSTRUCTIONS.md - The puzzle story and task description from Advent of Code
• part_1.py - Part 1 solution that prints only the raw answer (for easy piping)
• part_2.py - Part 2 solution that prints only the raw answer (for easy piping)
• test.py - Comprehensive test suite with example cases and edge cases
• input.txt - Personal puzzle input from Advent of Code
• README.md - Documentation of the approach, insights, and results

Running Solutions

Using mise (recommended)

# Run all parts for a day
mise run solve 1

# Run specific part
mise run solve 1 1      # Part 1
mise run solve 1 2      # Part 2

# Run tests
mise run test 1         # Test specific day
mise run test           # Test all days

Direct Python execution

Navigate to any day's directory and run:

# Navigate to a specific day (zero-padded)
cd days/01

# Run solutions (outputs raw answer)
python3 part_1.py
python3 part_2.py

# Copy answer directly to clipboard (macOS)
python3 part_2.py | pbcopy

# Run with custom input
python3 part_1.py custom_input.txt

# Run all tests
python3 test.py

Development Approach

Principles

1. Pure AI Development - No manual code edits; all changes through Claude Code
2. Test-Driven - Each solution includes comprehensive tests for edge cases
3. Clean Output - Solutions output only raw answers for easy automation
4. Self-Contained - Each day is independent with its own tests and documentation
5. Documented - Clear explanations of approach, insights, and key techniques

Workflow

1. Add the day's INSTRUCTIONS.md with the puzzle description
2. Use Claude Code to implement the solution
3. Create comprehensive tests covering edge cases
4. Document the approach and insights in the day's README
5. Verify the solution produces the correct answer
6. Move to the next day's puzzle

Code Quality

All solutions follow consistent patterns:

• Type hints for function parameters and return values
• Descriptive variable and function names
• Inline comments explaining non-obvious logic
• Separation of concerns (parsing, logic, I/O)
• Modular functions that can be tested independently

Setup

Prerequisites

1. Install mise - Tool version manager for Python and other dependencies

   curl https://mise.run | sh

   See mise.jdx.dev for other installation methods.

2. Install dependencies - All tools and versions are defined in mise.toml

   mise install

3. VSCode (recommended) - Open the project and search for @recommended in Extensions to install

   • Configuration is stored in .vscode/settings.shared.json
   • Extensions listed in .vscode/extensions.json

Quick Start

# Clone and setup
git clone https://github.com/balintant/AdventOfClaude-2025
cd AdventOfClaude-2025
mise install

# Run a day's solution
mise run solve 01

# Run tests for a specific day
mise run test 01

# Run all tests
mise run test

Technologies

• Tool Management: mise - See mise.toml for versions
• Language: Python 3.13 (standard library only, no external dependencies)
• IDE: VSCode with Claude Code extension
• AI Assistant: Claude (Sonnet 4.5)
• Testing: Custom test functions with assertions
• Linting: Ruff

Why This Matters

This project demonstrates:

• The capability of AI assistants to write production-quality code
• Effective human-AI collaboration through natural language
• The importance of good prompts and clear communication
• How AI can maintain consistency across a multi-day project
• The viability of "vibe coding" for solving algorithmic challenges

Progress

Day	Status	Part 1	Part 2
1	✅	1105	6599
2	✅	29940924880	48631958998
3	✅	17443	172167155440541
4	✅	1433	8616
5	✅	613	336495597913098
6	✅	5171061464548	10189959087258
7	✅	1698	95408386769474
8	✅	352584	9617397716
9	✅	4769758290	1588990708
10	✅	432	18011
11	✅	571	511378159390560
12	✅	414	⭐ (narrative only)

About Advent of Code

Advent of Code is an annual event featuring daily programming puzzles throughout December. Each day presents a two-part challenge that requires algorithmic thinking, problem-solving skills, and code implementation.

License

This project is for educational and demonstration purposes. Puzzle descriptions and stories are © Advent of Code.

Acknowledgments

• Eric Wastl for creating Advent of Code
• Anthropic for developing Claude and Claude Code
• The Advent of Code community for inspiration and support

---

Note: This repository showcases pure AI-assisted development. Every line of code, test, and documentation was generated by Claude Code without manual editing.