liveBook · Manning

Prompt engineering

As explained in Rearchitecting LLMs

Prompt engineering is the deliberate design and adjustment of instructions provided to a model to influence its outputs.

Prompt engineering deliberately designs detailed instructions or prompt chains to shape a model’s responses across tasks such as document analysis, report generation, coding, and translation. It can improve results without changing the model itself, but repeated use of a general model for specialized work can remain costly and fail to create meaningful differentiation. section 1.0 section 1.1.2

Prompt engineering is useful when:

  • A general-purpose model must be guided toward a particular response or workflow. section 1.0
  • A team needs to improve outputs without modifying model architecture or training a specialized model. section 1.1.2
  • Prompt chains organize multi-step behavior, while RAG supplies current or private information without transforming the model’s processing. section 1.1.2

Financial applications

A financial organization might prompt one general model to extract key metrics from thousands of legal documents and generate narrative client reports. Investment recommendations can still resemble competitors’ outputs when they use the same model and data. section 1.0 section 1.1.2

How Rearchitecting LLMs covers it

How deep it goes
Defines instruction adjustment and prompt chains, but stops short of changing underlying expertise or model architecture.
Examples
Uses financial-document extraction, client report generation, and investment recommendations as application examples.
What you can do after
Apply it through detailed instructions or prompt chains to shape responses while recognizing limits on specialization and differentiation.
How it connects
Connects to large language models, small language models, retrieval-augmented generation, fine-tuning, and model rearchitecting.

Also explained in

Compare how 51 books explain prompt engineering →

Related concepts

This explainer was generated by Manning's knowledge graph from the text of Rearchitecting LLMs. The section links point to the parts of the book it draws on.