Manifest AI

Today, we’re releasing Power Retention, a new architecture beyond Transformers. It enables LLMs to handle millions of tokens efficiently, unlocking long-context applications that were too costly before.