Anthropic discovers 'J-Space' inside Claude

Anthropic found a 'global workspace' within Claude where conscious-like reasoning occurs, with implications for AI safety. The discovery could reshape understanding of how large language models process information.