khalildh

khalildh / transversal-arc-solver

Public

ARC-AGI solver via Plücker geometry — 316 tasks solved with zero learning

84
11
100% credibility
Found Mar 31, 2026 at 84 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
C
AI Summary

A standalone program that solves same-size ARC-AGI grid transformation puzzles using geometric transversals and no machine learning training.

How It Works

1
🔍 Discover the Puzzle Solver

You stumble upon this clever tool online that cracks tricky grid puzzles without any learning or training.

2
📥 Grab the Puzzle Files

Download a collection of puzzle examples, each with a few input-output pairs and a mystery test grid to solve.

3
⚙️ Set Up the Solver

Get the single program file ready on your computer so it's all set to tackle puzzles.

4
🚀 Feed It a Puzzle

Point the solver at one puzzle file and watch it study the examples to uncover the hidden pattern.

5
⚡ Lightning Fast Check

It races through millions of possible answers in seconds, scoring them to find the perfect match.

6
📊 See the Results

Get back the top-ranked solution and check if it's number one – the best fit!

🎉 Puzzle Master Unlocked

Run it on hundreds of puzzles and celebrate solving most of them perfectly with pure clever math.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 84 to 84 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 transversal-arc-solver?

This C-based ARC-AGI solver tackles the ARC benchmark's grid transformation puzzles using Plücker geometry and transversals—no neural nets, no training data, zero learning required. Feed it ARC-AGI JSON files via CLI (`./arc_solver task.json` or `--all dir`), and it exhaustively scores candidates to find rank-1 solutions for same-size input/output grids. It nails 316 tasks from ARC-AGI training and eval sets, outputting solve status, timings, and predictions in seconds.

Why is it gaining traction?

Unlike ML-heavy ARC-AGI projects on GitHub chasing arc-agi-2 or arc-agi-3 with agents and Python deps, this pure C geometry approach delivers zero-shot wins on the arc-agi-benchmark without fitting parameters. Devs dig the raw speed—134M candidates scored in <1s via OpenMP—and its poetic take on solving ARC-AGI solutions geometrically, standing out amid poetiq-ai style solvers.

Who should use this?

ARC-AGI benchmark chasers building arc-agi-solver GitHub repos or arc-agi-python experiments. AI researchers probing non-learning paths to AGI puzzles, or geometry enthusiasts extending this to arc-agi-agent prototypes. Ideal for quick eval runs on same-size tasks without ML overhead.

Verdict

Grab it if you're deep in ARC-AGI GitHub scene—impressive 316 solved claim crushes brute force, compiles easily with LAPACK. But 84 stars and 1.0% credibility signal early-stage: thin docs, no tests, same-size only; fork and harden for production use.

(198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.