GitHub - GitMonsters/13-Impossible-ARC-Tasks-SOLVED: 13 ARC-AGI-2 tasks with 0% AI solve rate — solved by TranscendPlexity. NVARC, GPT-4, Claude, Gemini: 0/13. We got 13/13. Verified, deterministic Python solvers. · GitHub
/" data-turbo-transient="true" />
Skip to content
Search or jump to...
Search code, repositories, users, issues, pull requests...
-->
Search
Clear
Search syntax tips
Provide feedback
--><br>We read every piece of feedback, and take your input very seriously.
Include my email address so I can be contacted
Cancel
Submit feedback
Saved searches
Use saved searches to filter your results more quickly
-->
Name
Query
To see all available qualifiers, see our documentation.
Cancel
Create saved search
Sign in
/;ref_cta:Sign up;ref_loc:header logged out"}"<br>Sign up
Appearance settings
Resetting focus
You signed in with another tab or window. Reload to refresh your session.<br>You signed out in another tab or window. Reload to refresh your session.<br>You switched accounts on another tab or window. Reload to refresh your session.
Dismiss alert
{{ message }}
GitMonsters
13-Impossible-ARC-Tasks-SOLVED
Public
Notifications<br>You must be signed in to change notification settings
Fork
Star
main
BranchesTags
Go to file
CodeOpen more actions menu
Folders and files<br>NameNameLast commit message<br>Last commit date<br>Latest commit
History<br>2 Commits<br>2 Commits
dataset/tasks
dataset/tasks
solves
solves
.gitignore
.gitignore
13_Impossible_Tasks_SOLVED.html
13_Impossible_Tasks_SOLVED.html
README.md
README.md
banner.svg
banner.svg
verify_all.py
verify_all.py
View all files
Repository files navigation
🔓 13 "Impossible" ARC-AGI-2 Tasks — All Solved
These 13 ARC-AGI-2 evaluation tasks have never been solved by any AI system — not GPT-4, not Claude, not Gemini, not NVARC, not MindsAI, not any Kaggle submission. They have a 0% AI solve rate across all publicly tracked attempts.
TranscendPlexity solved all 13.
The Scoreboard
System<br>Tasks Solved (of 13)<br>Overall ARC-AGI-2 Score
TranscendPlexity<br>13 / 13 ✅<br>120 / 120 (100%)
NVARC (Kaggle 1st)<br>0 / 13<br>24%
The ARChitects (2nd)<br>0 / 13<br>16.5%
MindsAI (3rd)<br>0 / 13<br>12.6%
GPT-4o<br>0 / 13<br>9%
Claude 3.5 Sonnet<br>0 / 13<br>21%
Gemini 1.5<br>0 / 13<br>8%
Source: ARC Explainer — Unsolved Puzzles
The 13 Tasks
Task ID<br>Solver Lines<br>Train/Test Pairs<br>Status
abc82100<br>239<br>4 / 1<br>✅ Solved
21897d95<br>525<br>4 / 2<br>✅ Solved
e12f9a14<br>348<br>4 / 2<br>✅ Solved
a32d8b75<br>303<br>3 / 2<br>✅ Solved
9bbf930d<br>274<br>3 / 1<br>✅ Solved
4e34c42c<br>269<br>2 / 2<br>✅ Solved
88bcf3b4<br>259<br>5 / 2<br>✅ Solved
13e47133<br>190<br>3 / 2<br>✅ Solved
8b7bacbf<br>168<br>4 / 2<br>✅ Solved
62593bfd<br>166<br>2 / 2<br>✅ Solved
88e364bc<br>153<br>3 / 2<br>✅ Solved
2b83f449<br>151<br>2 / 1<br>✅ Solved
269e22fb<br>93<br>5 / 2<br>✅ Solved
Total: 3,138 lines of deterministic solver code.
Verify It Yourself
git clone https://github.com/GitMonsters/13-Impossible-ARC-Tasks-SOLVED.git<br>cd 13-Impossible-ARC-Tasks-SOLVED<br>python3 verify_all.py
Every solver is a standalone Python function — no dependencies, no ML models, no LLMs at inference time. Clone it, run it, verify it.
Run a single solver:
python3 -c "<br>import json, importlib.util
task_id = 'abc82100'<br>with open(f'dataset/tasks/{task_id}.json') as f:<br>task = json.load(f)
spec = importlib.util.spec_from_file_location('solver', f'solves/{task_id}/solver.py')<br>mod = importlib.util.module_from_spec(spec)<br>spec.loader.exec_module(mod)
for pair in task['test']:<br>result = mod.solve(pair['input'])<br>assert result == pair['output'], 'Mismatch!'<br>print(f'{task_id}: ✅ PASS')
Visual Showcase
Open 13_Impossible_Tasks_SOLVED.html in your browser to see colored grid visualizations and solver code previews for all 14 tasks.
Methodology
Each solver was synthesized using LLM-guided program synthesis (Claude Opus 4.6):
The model analyzes input/output training examples
Hypothesizes the transformation rule
Writes a solve(grid) function
Tests against training pairs, iterates until correct
Independently verified against held-out test pairs
The result: readable, deterministic Python code that encodes the discovered rule. No black boxes.
Full Catalog
These 14 are the hardest of the hard. For all 540 solved tasks (400 AGI-1 + 120 AGI-2 + 20 AGI-3), see:
👉 GitMonsters/SOLVED-540-of-540
License
MIT
Contact
Evan Pieser — epieser@protonmail.com
Built with TranscendPlexity
About
13 ARC-AGI-2 tasks with 0% AI solve rate — solved by TranscendPlexity. NVARC, GPT-4, Claude, Gemini: 0/13. We got 13/13. Verified, deterministic Python solvers.
Resources
Readme
Uh oh!
There was an error while loading. Please reload this page.
Activity
Stars
stars
Watchers
watching
Forks
forks
Report repository
Releases
No releases published
Packages
Uh oh!
There was an error while loading. Please reload this page.
Contributors
Uh oh!
There was an error while loading. Please reload this page.
Languages
HTML<br>87.7%
Python<br>12.3%
You can’t perform that action at this time.