How Mahesh Dives
Deep ?

Not like in traditional research.
Not just asking an AI to comment on a topic.
Instead, Mahesh engages the AI in a long conversation.

This is how Mahesh describes how he dives deep

I have a chat with an AI first. And this is how it typically and generally happens:

  1. I start with a Specific, Personal Observation or an Intuition or a Cheeky Question

I do not begin with a theory. I begin with something I had noticed or felt myself. I want to verify this observation/ intuition against what others had written.

  1. I use AI or an LLM as a Thinking Partner, Not an Authority

I do not ask the AI for answers. I ask them to reflect back what I am seeing. When they offer interpretations, I’d correct them—sometimes repeatedly—until they catch up with my thinking. Not the facts but the way of thinking. I treat AI as a mirror and a sounding board, not as a source of truth.

  1. I move from Concrete to Abstract and Back Again

I try to anchor every abstract claim in a concrete image or example or experience or established theory. This keeps the conversation grounded. Whenever the AI drifts into philosophy or drowns in recorded data, I pull them back to what a person can actually see, feel, or recognise without being subject to training or coaching various elements force upon us – without a washed brain.

  1. I Let the Argument Evolve in Layers

I do not present a finished thesis, because I don’t have one. Not yet. I build it step by step. Each layer emerging from the previous one. I do not force the structure. I let it grow.

  1. I Correct the AI Gently but Firmly

When they go in the wrong direction of thinking or reasoning or appraisal, I would say:  “No, to be honest.”; “Well, no, again.”; “Do you see it this way?”

I do not argue. I simply redirect. This keeps the conversation efficient and respectful. I treat AI as a tool that needed calibration, not an opponent to be defeated, or an oracle to be consulted or a mentor to listen to.

  1. I Knew When to Stop Adding

I sometimes go overboard in the chat for sake of the moment but then deliberately chose to leave it out of the final post. Because some insights are too deep or too personal for a public summary. I try to restrain myself—knowing what to exclude is as important as knowing what to include.

  1. I sometimes Ask for Multiple Versions of the Same Content

First, I may ask for a full, sophisticated document (for myrself, or for someone who wants the complete argument). Second, I may ask for a version a 12-year-old could understand (to test the clarity and core simplicity of the ideas). Third, I ask for a social media post for easy adult reading (to distill the essence for a wider audience). The AIs tell me that this multi-step distillation forced the ideas to become cleaner, sharper, and more portable. Each version revealed what was essential and what could be dropped.

  1. I Used AI to Externalize My Own Thinking

I do not ask AI to be creative, except in asking to do an image at times. I ask AI to reflect, summarise, translate, and format. I treat the AI as my external memory and my writing assistant. The ideas are mine. The structure emerges from my corrections. AI’s role is to say back to me what I had already said, in cleaner language, and to follow my leads.

  1. I am not Afraid to Trust My Own Intuition

Throughout the conversation, I try to trust my own felt sense of what is true or real to me—even when I cannot prove it. I am not afraid to say things like “I feel” and “my mind tells me”. I do not demand to “produce” evidence for every claim of mine. I allow intuition (which only I have got out of the two of us) to guide the inquiry, and only later do I ask for rational structure.

  1. I Decide When the Conversation is Complete

I end the substantive discussion on my own terms. Then I thank the AI and ask for the final summary. I try not to let the conversation drag on but close it cleanly.

Role of AI in this: This is not a technique that requires artificial intelligence. It is a technique for thinking clearly with any tool, any partner, or even just a notebook. The AI made it faster and more conversational. The method is simply starting small, correcting often, distilling repeatedly, and trusting my own mind.

How Mahesh Dives
Deep ?

Not like in traditional research.
Not just asking an AI to comment on a topic.
Instead, Mahesh engages the AI in a long conversation.

This is how Mahesh describes how he dives deep

I have a chat with an AI first. And this is how it typically and generally happens:

  1. I start with a Specific, Personal Observation or an Intuition or a Cheeky Question

I do not begin with a theory. I begin with something I had noticed or felt myself. I want to verify this observation/ intuition against what others had written.

  1. I use AI or an LLM as a Thinking Partner, Not an Authority

I do not ask the AI for answers. I ask them to reflect back what I am seeing. When they offer interpretations, I’d correct them—sometimes repeatedly—until they catch up with my thinking. Not the facts but the way of thinking. I treat AI as a mirror and a sounding board, not as a source of truth.

  1. I move from Concrete to Abstract and Back Again

I try to anchor every abstract claim in a concrete image or example or experience or established theory. This keeps the conversation grounded. Whenever the AI drifts into philosophy or drowns in recorded data, I pull them back to what a person can actually see, feel, or recognise without being subject to training or coaching various elements force upon us – without a washed brain.

  1. I Let the Argument Evolve in Layers

I do not present a finished thesis, because I don’t have one. Not yet. I build it step by step. Each layer emerging from the previous one. I do not force the structure. I let it grow.

  1. I Correct the AI Gently but Firmly

When they go in the wrong direction of thinking or reasoning or appraisal, I would say:  “No, to be honest.”; “Well, no, again.”; “Do you see it this way?”

I do not argue. I simply redirect. This keeps the conversation efficient and respectful. I treat AI as a tool that needed calibration, not an opponent to be defeated, or an oracle to be consulted or a mentor to listen to.

  1. I Knew When to Stop Adding

I sometimes go overboard in the chat for sake of the moment but then deliberately chose to leave it out of the final post. Because some insights are too deep or too personal for a public summary. I try to restrain myself—knowing what to exclude is as important as knowing what to include.

  1. I sometimes Ask for Multiple Versions of the Same Content

First, I may ask for a full, sophisticated document (for myrself, or for someone who wants the complete argument). Second, I may ask for a version a 12-year-old could understand (to test the clarity and core simplicity of the ideas). Third, I ask for a social media post for easy adult reading (to distill the essence for a wider audience). The AIs tell me that this multi-step distillation forced the ideas to become cleaner, sharper, and more portable. Each version revealed what was essential and what could be dropped.

  1. I Used AI to Externalize My Own Thinking

I do not ask AI to be creative, except in asking to do an image at times. I ask AI to reflect, summarise, translate, and format. I treat the AI as my external memory and my writing assistant. The ideas are mine. The structure emerges from my corrections. AI’s role is to say back to me what I had already said, in cleaner language, and to follow my leads.

  1. I am not Afraid to Trust My Own Intuition

Throughout the conversation, I try to trust my own felt sense of what is true or real to me—even when I cannot prove it. I am not afraid to say things like “I feel” and “my mind tells me”. I do not demand to “produce” evidence for every claim of mine. I allow intuition (which only I have got out of the two of us) to guide the inquiry, and only later do I ask for rational structure.

  1. I Decide When the Conversation is Complete

I end the substantive discussion on my own terms. Then I thank the AI and ask for the final summary. I try not to let the conversation drag on but close it cleanly.

Role of AI in this: This is not a technique that requires artificial intelligence. It is a technique for thinking clearly with any tool, any partner, or even just a notebook. The AI made it faster and more conversational. The method is simply starting small, correcting often, distilling repeatedly, and trusting my own mind.