3
Prompting DeepSeek
If you’ve ever moved from writing imperative code, manually telling the Domain Object Model (DOM) how to update a web page step-by-step, to a declarative framework such as React, you know the required mental shift. You stop giving meticulous instructions and start describing the desired outcome, trusting the system to figure out the how. The same principle applies when moving from traditional LLMs such as Mistral 7B or Llama 3.3 8B to DeepSeek’s reasoning engines. The muscle memory you’ve developed for carefully guiding a model’s thought process might actually work against you with an engine that’s built to reason on its own.
This chapter is all about navigating those differences, from the headline-grabbing quirks (such ...
Become an O’Reilly member and get unlimited access to this title plus top books and audiobooks from O’Reilly and nearly 200 top publishers, thousands of courses curated by job role, 150+ live events each month,
and much more.
Read now
Unlock full access