Knowledge cutoff

A knowledge cutoff is the last date covered by a model's training data. Anything that happened after that date is missing from what the model learned.

Models are often used for months after their cutoff, so the gap between the cutoff and the present keeps growing.

• The model cannot know recent changes, such as a new price or a new contact.
• The model never saw private data, such as a customer's notes, so it cannot answer about it from training alone.

• Harny gives each person a private memory that the app's AI can read, so the app can supply facts that came after the cutoff.
• The app sends private data to memory with the Memory API, and the AI can read it at answer time.

No. Every model has a cutoff. Memory lets an app supply newer and private facts at answer time.