PulseAugur
EN
LIVE 17:52:15

Open-source tool Corteza scores 98.6% on ARC-AGI-3 with Claude Opus 5

By PulseAugur Editorial · [1 sources] ·

An open-source tool named Corteza, developed by Troy, achieved a score of 98.6% on the ARC-AGI-3 benchmark. This achievement was accomplished using Claude Opus 5 and involved completing all 183 levels across 25 public games. The Corteza harness is designed to write and reuse functions within a persistent R workspace, allowing data and functions to persist through conversation compaction. AI

IMPACT Demonstrates advanced reasoning capabilities of LLMs on complex benchmarks, potentially influencing future AI development.

RANK_REASON Research benchmark result using an LLM. [lever_c_demoted from research: ic=1 ai=1.0]

Read on Mastodon — mastodon.social →

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →

Open-source tool Corteza scores 98.6% on ARC-AGI-3 with Claude Opus 5

How we ranked this

Signal score
16 / 100
Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket
Tool
Research benchmark result using an LLM. [lever_c_demoted from research: ic=1 ai=1.0]
Source corroboration
Single-source cluster
Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics
product, model release
Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance
High
Clearly on-topic for AI-industry coverage.
Story freshness
Breaking (< 6h)
Fresh story with cross-source coverage still developing. Ranking may shift as more sources report.

Full methodology in our editorial standards.

COVERAGE [1]

  1. Mastodon — mastodon.social TIER_1 English(EN) · troyhernandez ·

    Hi, I’m Troy. I build open-source tools in R. My agent harness, corteza, scored 98.6% on ARC-AGI-3’s 25 public games with Claude Opus 5 at xhigh, completing all

    Hi, I’m Troy. I build open-source tools in R. My agent harness, corteza, scored 98.6% on ARC-AGI-3’s 25 public games with Claude Opus 5 at xhigh, completing all 183 levels. The harness writes and reuses functions in a persistent R workspace. Functions and data survive conversatio…

    LINKS cornball.ai/…/corteza-arc-agi-3

RELATED ENTITIES

RELATED TOPICS