• AIpeeps
  • Posts
  • Learning Loop ♾️ 10-2025

Learning Loop ♾️ 10-2025

Your bi-weekly dose of learning AI, without the hype.

Author

Gaurav Jain
November 29, 2025 • Estimated Reading Time: 6 minutes

Hello peeps,

Hello everyone, and a very warm welcome to the 10th edition of AIpeeps!

What we’re covering today:

  • Prompt Refinement Loop: and why it gives cleaner, smarter results

  • NotebookLM’s Studio: videos, podcasts, reports, mind maps, flashcards, quizzes

  • Google’s Setings to stop your data-for-AI setting that sent my Gmail straight back to 2010

  • Semantic Search — how it changes results completely

💡3 Curious things I learnt

1. Iterative Prompt Rewriting (aka Prompt Refinement Loop)

Letting ChatGPT ask clarifying questions first is one of the smartest ways to handle complex tasks.

But here’s the twist — answering those questions inside the chat often clutters the thread (what advanced prompt engineers call “chat pollution”).

Here’s a clean strategy to get better results 👇

  • Initial Prompt: Ask ChatGPT to first pose clarifying questions so it can do a good job.

  • Receive Questions: Review the questions it asks.

  • Answer in the Prompt: Go back and edit your original prompt to include your answers.

  • Resubmit: Submit the refined prompt and continue from there.

Why This Works

✅ Improved Quality: You give all details upfront → AI delivers sharper results.
✅ Avoids Confusion: Less back-and-forth, fewer missteps.
✅ Enhanced Reasoning: The AI thinks better when your answers are structured.

PS- Till now, I was also stuck in “First Ask Then Answer”. FATA = Ask clarifying questions → User answers → Model proceeds.

2. NotebookLM — Your Research-to-Content Workspace (Step-Wise Breakdown)

In short, you can also consider it as the pro version of ChatGPT “Study and Learn.”

Go to the NotebookLM website; create an account using Gmail.

After creating your account, click “Create New” to start your notebook.

After that, add PDFs, Docs, Google Drive files, web pages, or YouTube videos (auto-transcribed)

The interface is organized into three panels as shown in image:

  • Sources on the left

  • Chat area in the center

  • AI tools (video, podcast, mind map, reports, flashcards, quizzes) on the right

Other Highlights:

  • Use “Discover” to instantly import high-quality links from the web

  • It only answers from your chosen sources — with inline citations

  • NotebookLM Studio can auto-convert your notes into videos, mind maps, or reports — no complex prompting!

3. Google’s Data for AI Training

This one’s been all over social media lately — people are saying if you untick “Smart feature” under setting, you can stop Google from using your data for AI training.

So I tried it.

😂 Big mistake — it sent my Gmail straight back to 2010.

Smart -Turn Off

Smart - Turn On

No smart features, no categorization, parcel updates, predictive typing etc. Definitely not for me.

From what I can see, the only useful toggle is the one in Workspace settings (below “Smart features”) — that doesn’t really change my day-to-day use.

PS - I still feel, you just can’t save your data from google; they just know everything about you.

🕵🏻Decoding the Jargon

Semantic search is an advanced method of information retrieval designed to understand the user’s intent, the context of the query, and the meaning behind the terms used—rather than simply matching keywords on a page.

This makes search systems more accurate, intuitive, and closer to how humans naturally ask questions.

A Knowledge Graph (KG) plays a central role in enabling semantic search. It provides a structured way to represent real-world concepts and how they relate to each other.

Foundation of Semantic Search

Semantic search operates by combining several advanced technologies:

Natural Language Processing (NLP)

Helps the system parse, interpret, and understand human language—including grammar, phrasing, synonyms, and context.

Machine Learning (ML)

Allows the system to learn patterns, relationships, and meanings from data over time, improving search quality and relevance.

Knowledge Graphs (KGs)

Provide a structured representation of real-world entities (people, places, events, products, concepts) and the relationships between them.
This interconnected structure helps the system interpret meaning and context more accurately.

Keyword Search vs. Semantic Search

  • Keyword search: Looks for exact word matches in documents; meaning is not considered.

  • Semantic search: Attempts to understand what the user is actually asking, even if the exact words are not used.

Semantic search therefore delivers more relevant and context-aware results.

Try it Out this week:

✔️ Try making a notebook on NotebookLM

✔️ Use a Iterative Prompt Rewriting to sharpen your prompt

Share the love, flex what you learned, and let’s grow this convo. 🚀

Please feel free to share any questions or topics you're particularly curious about.

Hope you learned something new today!

Till next time,

Reply

or to participate.