Make Fabrication Unrepresentable
A practical approach to preventing AI assistants from fabricating numbers by making model outputs cite or compute values instead of asserting them.
llm
3 articles
The Tools You Build Are More Important Than The Tools You Use
Why developers succeeding with AI coding assistants stop searching for perfect configurations and start building their own tools.
The Language Model Is Just Another User
A CQRS-inspired approach to integrating LLMs: treat the language model as a user sending commands, not an internal system component.