Sakana AI unveils My Fish to reveal chatbot memories

Analyst assessment: My Fish addresses a practical question for persistent assistants: what information is shaping an answer? Sakana's interface exposes remembered facts, their use over time and relationships created when memories are used together. That gives users a way to inspect personalization rather than simply accept it. For Race to AGI readers, the significance is architectural and operational: an assistant's behavior depends on its surrounding memory system as well as the underlying model. ([sakana.ai](https://sakana.ai/my-fish/))

The competitive opportunity is to make continuity understandable. A useful evaluation would ask whether users can identify mistaken memories, understand unexpected responses and decide when to turn persistence off. Visibility alone does not establish correct retrieval, reliable reasoning or comprehensive privacy protection. Nor does this release demonstrate a new general capability. It does, however, offer a concrete interface for studying how people supervise an assistant that carries information across conversations. My assessment is that this is an incremental product advance with potential value for agent oversight, rather than evidence that the AGI timeline has accelerated. The strongest follow-up would measure whether the interface actually helps users catch and manage memory-related errors. ([sakana.ai](https://sakana.ai/my-fish/))