Gemini: Deep Research
A first experiment
I accessed Gemini Deep Research for the first time this morning to see what it could do. I was pretty impressed.
I asked it very simply "what is the optimum way to structure a prompt for a LLM?" and off it went, explaining how it would conduct the research, the steps it would follow and checking I was happy with that approach.
It completed its research, correctly citing sources, in under 4 minutes and then generated a report that I was able to check and add to with my own findings. It also created exportable tables for guidance.
I was even able to get a suggested acrostic to frame effective prompts:
CONTEXT
Clear: Be clear and specific in your instructions.
Objective: Have a well-defined objective in mind.
Nuanced: Provide relevant context and background information.
Targeted: Specify the desired format of the output.
Examples: Use examples to guide the LLM.
Xplicit: Explicitly state the request or task.
Tailored: Adjust the prompt based on the LLM’s response.