I’ve heard a lot this approach on X but hadn’t seen somebody break it down like this. I also appreciate the way you’re weaving AI Meta Cognition into the practical AI assisted development and pointing to the broader social considerations.
I think AI can democratize opportunity but only if the early adopters choose to onboard folks who would otherwise not have the inclination to use these tools!
That’s why I’m teaching impact founders in the global south how to build products with AI for free, to give back from the opportunity I gained simply by being born in the US during the .com boom
Feels like this is introducing a new paradigm on top of AI-Assisted coding which is adding optimized mechanisms to ground the software changes in human understanding.
When you consider that the amount invested today in the AI race dwarfs what was invested in the moon landing race, it makes you wonder…
Especially when China is accomplishing very close to the same with their AI work at 10x cheaper costs.
That’s one reason I believe that OpenAI is headed either for absorption by a bigger player or bankruptcy. Its business model is unsustainable.
The other big U.S. companies have the resources to absorb whatever big bet they put into AI but not OpenAI.
To your point about research and new discoveries, as we learn how to combine the various tools in new and powerful ways, it will provide us with a powerful architecture for learning. Already is doing that in many ways. It should continue improving throughout 2026.
Claude Code that became Cowork is one example of this effect at work.
OpenAI had this huge lead and they basically dropped the mother of all balls! 🤦♂️
Yes NBLM is not well understood by the majority but has just so many use cases. In Feb 2025 I had sat my two girls down, both in university, and told them that they were not using AI to their advantage in their studies and then introduced them to NBLM.
It keeps getting better and better.
Today, it is a game changer in so many domains. Especially with the 300 sources limit to the Pro plan.
You also have an excellent point on Google’s ecosystem.
I recently discovered Rube.app from Composio. It can turn Claude into a massive ecosystem connecting over 600 apps together. Works for the LLM and Claude Code.
It is very good at developing what it calls recipes and when it encounters a bottleneck or issue, to find a workaround.
I also came across a workflow with programmers use Claude code to generate the initial architecture and then Kimi+ to do the iterations / bug fixes cause it is much cheaper.
I think we are only at the beginning of tool synergies that are going to 10x productivity.
I’m really glad to hear your encouraging and introducing your daughters to NoteBookLM and AI tools at large.
Sure these things can be misused and harm learning. But anything, and I mean everything. Has positive and negative use cases
Used correctly AI has the power to super charge learning. Your encouraging them to use it to help them learn is one of the best things you can do to prepare them for the coming wild future ahead of us.
Like it or not, and I agree there are harms that come from AI. This is the future, to not learn to use AI is like continuing to ride horses when cars are available.
It’s here and it’s only going to integrate more into the world . Cutting it out of society is not realistic. Nobody can stop the flow of trillions of dollars and the extreme research interest.
The only thing anyone can do is learn to advantage the positive use cases
Yeah, my youngest is anti-technology in general like many of her generation. I did my best to explain to them that yes, AI can be used to cheat but that doesn’t make AI evil. It makes anyone doing that just plain lazy and lacking in imagination.
I started my introduction by telling them “I promise you that you are using AI wrong and that you could be getting way more out of it than you realize.” My older daughter turned to my younger one and said, “Most kids are getting chastised for making poor lifestyle choices and we are getting it over our use of AI.” Lol.
Even my anti-tech kid had to admit that NBLM was useful for University studies.
You've added incredible context to this comment thank you for sharing. Exactly and I think these are the sorts of AI tools that gives us some hope that this can improve eventually academics, research and even the scientific method and the speed of new discoveries.
Google has some potential to make breakthroughs now for AI in science that might be one of the more tangible AI for good capabilities they can offer the world.
When I think of how good the openweight models of China have become like Qwen-3 Max Thinking, we are reaching a point where the AI startups that will be founded will be considerably better than the ones that were even possible two years ago.
All of this does scale at the frontier towards something better on a fundamental level. If we can reduce the cost of research and development and enable new kinds of innovation, all of civilization are the winners in the end.
Thank you for your comment Joseph. NBLM has to be one of the most improved AI tools of 2025. I can only imagine Claude code will make similar improvements in terms of adding useful features and capabilities in 2026.
Since Alphabet is an investor in Anthropic, I think we can expect some interoperability between these two ecosystems.
A bit off topic but AI overviews is increasing traffic to YouTube in a significant way. Google's full stack and ecosystem advantage is often underestimated. Anthropic on the other hand is going to be so laser focused on Enterprise AI, it's going to add so much value and revenue by doing what it does better than everyone else.
Using mindmaps made in notebook is a smart idea. I use words and it's often not what I wanted, this is a really interesting alternative. Thanks for taking the time.
Glad you liked it! I do really like the mind maps for complex topics, the visual spread of it, to see where everything connects to everything else is a really good supplement to anything you’re trying to learn
This piece really made me think! So good to see you building on thoes earlier insights.
Let me know if any of the guides or articles at the bottom were useful, and if so, which ones.
I’ve heard a lot this approach on X but hadn’t seen somebody break it down like this. I also appreciate the way you’re weaving AI Meta Cognition into the practical AI assisted development and pointing to the broader social considerations.
I think AI can democratize opportunity but only if the early adopters choose to onboard folks who would otherwise not have the inclination to use these tools!
That’s why I’m teaching impact founders in the global south how to build products with AI for free, to give back from the opportunity I gained simply by being born in the US during the .com boom
Feels like this is introducing a new paradigm on top of AI-Assisted coding which is adding optimized mechanisms to ground the software changes in human understanding.
Awesome article!
I gotta try some of the flows, amazing article Michael! Just gave you a follow!
Love the comprehensive workflow using NotebookLM to understand the code. I had not thought to use NotebookLM this way before.
What a game changer!
When you consider that the amount invested today in the AI race dwarfs what was invested in the moon landing race, it makes you wonder…
Especially when China is accomplishing very close to the same with their AI work at 10x cheaper costs.
That’s one reason I believe that OpenAI is headed either for absorption by a bigger player or bankruptcy. Its business model is unsustainable.
The other big U.S. companies have the resources to absorb whatever big bet they put into AI but not OpenAI.
To your point about research and new discoveries, as we learn how to combine the various tools in new and powerful ways, it will provide us with a powerful architecture for learning. Already is doing that in many ways. It should continue improving throughout 2026.
Claude Code that became Cowork is one example of this effect at work.
OpenAI had this huge lead and they basically dropped the mother of all balls! 🤦♂️
Yes NBLM is not well understood by the majority but has just so many use cases. In Feb 2025 I had sat my two girls down, both in university, and told them that they were not using AI to their advantage in their studies and then introduced them to NBLM.
It keeps getting better and better.
Today, it is a game changer in so many domains. Especially with the 300 sources limit to the Pro plan.
You also have an excellent point on Google’s ecosystem.
I recently discovered Rube.app from Composio. It can turn Claude into a massive ecosystem connecting over 600 apps together. Works for the LLM and Claude Code.
It is very good at developing what it calls recipes and when it encounters a bottleneck or issue, to find a workaround.
I also came across a workflow with programmers use Claude code to generate the initial architecture and then Kimi+ to do the iterations / bug fixes cause it is much cheaper.
I think we are only at the beginning of tool synergies that are going to 10x productivity.
I’m really glad to hear your encouraging and introducing your daughters to NoteBookLM and AI tools at large.
Sure these things can be misused and harm learning. But anything, and I mean everything. Has positive and negative use cases
Used correctly AI has the power to super charge learning. Your encouraging them to use it to help them learn is one of the best things you can do to prepare them for the coming wild future ahead of us.
That’s awesome of you!
Like it or not, and I agree there are harms that come from AI. This is the future, to not learn to use AI is like continuing to ride horses when cars are available.
It’s here and it’s only going to integrate more into the world . Cutting it out of society is not realistic. Nobody can stop the flow of trillions of dollars and the extreme research interest.
The only thing anyone can do is learn to advantage the positive use cases
Agreed. Pandora’s box has been opened, for better or for worse.
Some great things will come from AI and hopefully, we will learn to control and mitigate the bad.
Yeah, my youngest is anti-technology in general like many of her generation. I did my best to explain to them that yes, AI can be used to cheat but that doesn’t make AI evil. It makes anyone doing that just plain lazy and lacking in imagination.
I started my introduction by telling them “I promise you that you are using AI wrong and that you could be getting way more out of it than you realize.” My older daughter turned to my younger one and said, “Most kids are getting chastised for making poor lifestyle choices and we are getting it over our use of AI.” Lol.
Even my anti-tech kid had to admit that NBLM was useful for University studies.
You've added incredible context to this comment thank you for sharing. Exactly and I think these are the sorts of AI tools that gives us some hope that this can improve eventually academics, research and even the scientific method and the speed of new discoveries.
Google has some potential to make breakthroughs now for AI in science that might be one of the more tangible AI for good capabilities they can offer the world.
When I think of how good the openweight models of China have become like Qwen-3 Max Thinking, we are reaching a point where the AI startups that will be founded will be considerably better than the ones that were even possible two years ago.
All of this does scale at the frontier towards something better on a fundamental level. If we can reduce the cost of research and development and enable new kinds of innovation, all of civilization are the winners in the end.
Thank you for your comment Joseph. NBLM has to be one of the most improved AI tools of 2025. I can only imagine Claude code will make similar improvements in terms of adding useful features and capabilities in 2026.
Since Alphabet is an investor in Anthropic, I think we can expect some interoperability between these two ecosystems.
A bit off topic but AI overviews is increasing traffic to YouTube in a significant way. Google's full stack and ecosystem advantage is often underestimated. Anthropic on the other hand is going to be so laser focused on Enterprise AI, it's going to add so much value and revenue by doing what it does better than everyone else.
Using mindmaps made in notebook is a smart idea. I use words and it's often not what I wanted, this is a really interesting alternative. Thanks for taking the time.
Glad you liked it! I do really like the mind maps for complex topics, the visual spread of it, to see where everything connects to everything else is a really good supplement to anything you’re trying to learn