theredbluepill

theredbluepill / arc-interactive

Public

A collection of community game environments for the ARC-AGI-3 benchmark.

docs.arcprize.org
19
1
100% credibility
Found Mar 24, 2026 at 19 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Python
AI Summary

A community collection of interactive puzzle games designed to test AI reasoning, planning, and control skills.

How It Works

1
🔍 Discover puzzle games for AI

You find a collection of clever puzzles that humans solve easily but challenge AI to think and plan.

2
📥 Get the games ready

Download everything and prepare it on your computer in moments—no coding needed.

3
📋 Browse the game list

See all the interactive games available, like Sokoban or memory match, with previews.

4
🎮 Play like a human

Use arrow keys and space to move and act in a game window, learning the rules by doing.

5
🤖 Test an AI player

Watch a simple random player try the puzzle, or connect online to see real scores on a leaderboard.

6
Go online or stay local
🚀
Online leaderboard

Submit scores to compete with top AIs worldwide.

🏠
Local practice

Practice offline with full control and no waiting.

🏆 Master the challenge

You train AIs, beat puzzles, or even add your own game to the collection—fun for everyone!

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 19 to 19 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is arc-interactive?

Arc-interactive is a Python collection of 249 community-built game environments for the ARC-AGI-3 benchmark, focusing on interactive puzzles that test AI reasoning, planning, and control. Developers run games via a simple CLI like `uv run python run_game.py --game ez01 --mode human` for offline play with WASD controls, random agents, or online leaderboard submission. It solves the need for diverse, human-solvable tasks beyond static grids, with tutorial levels and competition mode matching official rules.

Why is it gaining traction?

Unlike official ARC-AGI-3's limited set, this packs 200+ varied games into one repo for quick agent testing and generalization checks, all playable locally without internet. Human mode lets you explore win conditions hands-on, while LLM authoring skills speed up custom game creation. Benchmarks integrate with Kaggle for easy model eval on tasks like Sokoban or survival.

Who should use this?

AI researchers benchmarking LLMs on planning-heavy games, ARC Prize competitors needing offline prototyping, or devs building interactive agents frustrated by template puzzles. Ideal for teams evaluating skill-acquisition on unseen tasks before leaderboard runs.

Verdict

Grab it for ARC-AGI-3 prototyping—solid CLI and human play make iteration fast, despite 19 stars and 1.0% credibility signaling early-stage maturity. Docs are clear with quickstarts; add tests for production confidence. Worth forking if you're in AGI evals.

(198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.