How to Learn Something New Efficiently

How to Learn Something New Efficiently

This is a method I use for quickly getting up to speed on a new knowledge area — broad or deep, either works. It applies to any domain, but I find it especially useful for technical or conceptual material (think: monetary concepts, macro theory, policy frameworks). I recommend Gemini for text-based learning and ChatGPT for voice practice. For multilingual people, I also recommend conducting all conversations in English — in my experience, AI performance on technical material drops noticeably in other languages.

The process has four parts.

Part 1 — Map the Outline

Start by asking the AI to give you a structured overview of the topic. The goal is not to understand everything yet — it is to know what exists so you can navigate it.

Ask something like: “What are the main responsibilities and areas of influence of [institution/concept]? Give me a structured outline.”

This step also serves as a quick orientation to the scope of what you are learning, which is useful context before going deeper.

Part 2 — Learn Step by Step with Gemini

Once you have the outline, open Gemini and use the Learn option. Give it the background from Part 1 and the specific area you want to learn, then select “Help me learn.”

From here, treat it like a personal tutor. Follow its prompts one step at a time.

A few things that make this step work well:

  • Answer every question it poses. Do not skip them. Formulating an answer — even an uncertain one — is what builds familiarity.
  • Ask follow-up questions on each answer. This both deepens your understanding and surfaces the kinds of follow-up questions you might face in a real conversation.
  • If the conversation runs long and the AI loses track of earlier context, just open a new chat and pick up from the specific sub-topic you want to continue with.

Example follow-up prompt: “What theoretical frameworks does the surveillance function of monetary institutions rely on, and why?”

Part 3 — Practice with Scenarios

After Part 2, open a fresh conversation — no Learn mode. Ask the AI to generate scenario-based questions you might face in a serious discussion of this topic:

“Give me a set of questions someone might ask in an interview or technical discussion about [topic].”

Then work through them one by one. For each answer the AI gives, push back and ask follow-ups. In my experience, AI handles conceptual material at roughly undergraduate difficulty without hallucinating, so this is a reliable way to build fluency.

Example scenario prompt: “For a developing economy, how would you assess whether its debt level is sustainable?”

Part 4 — Practice Out Loud with Voice

This step is specifically for people preparing technical questions for an interview.

Reading and typing are not the same as speaking. The last step is to open ChatGPT’s voice mode and simulate a spoken conversation on the topic.

I recommend two rounds:

  1. You as interviewer, AI as respondent — ask it the questions and evaluate its answers. This gives you a model for how the material should be explained.
  2. AI as interviewer, you as respondent — now you answer, and it pushes back.

ChatGPT’s voice mode is significantly better than Gemini’s for this purpose in my experience.