
Tech
The Model Wasn't Stupid. It Was Blind: How a Four-Sentence Prompt Took Claude to a Perfect 100 on ARC-AGI-3
One-line: On October 1, MIT's Kaiming He team published VISTA — a visual harness with a lossless photo-album memory that lifted Claude Opus 5.0's Relative Human Action Efficiency on ARC-AGI-3 from 40.68 to a perfect 100.00 , clearing all 25 public games with 57.4% fewer actions than first-time human players. No new model. No training. Four sentences of prompt.
ARC-AGI-3 is the cruel exam of the AI world. It hands an agent an unfamiliar video game — grid-world puzzles, checkers varian...
Read the full discussion on Dev.to
This article was aggregated from Dev.to. Click to join the conversation.
View on Dev.to