How to use AI for Cutting Edge Research
Featuring insights from Andrew Ng, Terence Tao, Yann LeCunn, and many other top researchers on how they use AI for better gains
Every month, the Chocolate Milk Cult reaches over a million Builders, Startup Founders, Investors, Policy Makers, Leaders, and more. If you’d like to meet other members of our community, please fill out this contact form here (I will never sell your data nor will I make intros w/o your explicit permission)- https://forms.gle/Pi1pGLuS1FmzXoLr6
There is a widening chasm in the world of Artificial Intelligence.
On one side, we have formal productivity studies—like the recent METR report—concluding that AI provides “zero meaningful speedup” for experts. On the other side, we have the world’s most elite researchers, from Fields Medalist Terence Tao to AI pioneer Andrew Ng (I’m specifically focusing on the guys who have no conflict of interest here), claiming that AI has fundamentally transformed their ability to explore “crazier” ideas at a higher velocity.
So, who is lying?
Neither. The problem is that most people are measuring the wrong variable. They are treating AI as a replacement for the tasks they are already good at—the exact area where the tool provides the least amount of leverage.
If you’ve been using AI to “write faster” and feeling underwhelmed, you’re missing the real game. For this Chocolate Milk Cult exclusive, we dug deep into our own researchers+their workflows, cross-checked multiple sources, and finally spoke to several of the leading researchers in the world (including one of Nvidia’s lead researchers who directly reports to Jensen; trying to have them on the livestream soon) to understand how they are using AI to do cutting-edge research:
The “Local Maximum” Trap: Why most people use AI in a way that actually introduces more friction, and the mental shift required to fix it.
The Scout vs. The Strategist: The specific “Periphery” framework that separates a mediocre AI prompt from a breakthrough research insight.
Capturing Superlinear Gains: How to identify and automate the “dead time” that traditional productivity surveys completely ignore.
The Mediocrity Arbitrage: A counter-intuitive strategy for using “average” AI outputs to create “elite” end-results.
The Next Bottleneck: Why the future of AI leverage has nothing to do with better generation—and what you should be building instead.
This article is written to be useful to more people than pure researchers. Whether you are an engineer, a founder, or a creative, you will find principles that apply to managing complex projects, building products, or solving non-routine problems with AI.
To access the full article—and all premium breakdowns going forward/written prior—upgrade to a premium subscription below.
If you believe deep insight deserves support, become a premium subscriber to allow me to keep doing the same.
Flexible pricing available—pay what matches your budget here.
Most companies offer learning or professional development budgets. You can expense this subscription using the email template linked here.
