How I Use AI to Write 3 Substacks and Run 4 Companies

Every word I publish online or privately to my partners gets polished by me.

…but I outline, lateral think, counter-point, condense, remix and fuse with AI.

I use it across my personal, professional and creative life.

24/7 — I even use multiple AI tools. We’ll get to that in detail.

I use this process to:

• write Python scripts to automate / explore data

• plan expansions to one of my businesses

• study opportunities for investment or development

As you can see I use both ChatGPT by OpenAI and Gemini from Google.

The best part about Gemini (that sets it apart from ChatGPT):

1. Deterministic skills → math, coding, structured work like engineering

2. Probabilistic skills → generating text

ChatGPT is built on a probability engine - it’s basically a number calculator that outputs words.

It doesn’t do great at math. Large Language Model, right?

ChatGPT is good at talking but our universe is written in code. Math is code. Business, engineering, everything is structured and seeking higher structure.

Gemini has a unique structure designed by Google.

It can reason deterministically (solve math, generate code, etc..) and work probabilistically to create a text ‘envelope’ around whatever you are creating.

Probabilistic models can handle novel and unexpected inputs by adjusting their responses based on statistical patterns. They can generate a variety of responses for the same input, showcasing creativity and flexibility.

Probabilistic models can express uncertainty and quantify confidence in their responses, providing more nuanced output.

Really exceptional mathematicians use intuition when they solve novel problems. Being able to give AI a brain that leverages probabilistic and deterministic reasoning will open the door for intuitive solutions to come from synthetic sources.

‘Growing’ an Article

One of my favorite methods with AI is to grow an article.

The Skeleton - Outline

I start by feeding ideas and angles into the AI, and I ask for 3 outlines to be generated.

Requesting multiple responses (vs a single) gives you “more bites at the apple”. Often I take my favorite elements of each outline and then I cobble them together.

The Meat - Bullet Expansions

Once my ideal outline is complete I start a new thread to avoid context contagion from the input I fed into the AI to generate the outline.

A great starting point is to ask the AI to draft 5 bullets to summarize each step in the outline.

This will suddenly start to add meat to the outline.

Your prompt technique will greatly determine the value of the output you generate. Consider fine tuning your prompts with phrases including:

• “use a first-person perspective”

• “adopt the voice of Mark Twain, write humorously with modern language”

• “make sure the bullets proceed chronologically / logically / order of importance, etc…“

The Spirit - Finishing Touches

At this point we have a body of text that is logically constructed.

We don’t have an article of any value unless we have all of this:

1. Headline: This is the first element that catches the reader's attention. A good headline should be captivating, informative, and reflective of the content.

2. Narrative Flow: The content should follow an engaging sequence, often beginning with an introduction, moving through various points or arguments, and culminating in a conclusion. This flow helps maintain coherence and clarity.

3. Lead or Hook: The opening paragraph or sentence, known as the lead, should hook the reader. It might pose a question, present a surprising fact, or set a scene to draw readers into the article.

4. Bridge: This element connects the introduction to the main body of the article. It provides a smooth transition, maintaining the reader's interest and leading them naturally into the deeper content.

5. Subheadings: These help break the text into manageable sections. Subheadings guide the reader through the article and make the content more scannable, especially important in digital formats.

6. Evidence and Examples: To reinforce points made, it's important to include relevant examples, case studies, or data. This not only adds credibility but also helps the reader understand and relate to the content.

7. Visual Elements: In digital formats, incorporating images, videos, or infographics can greatly enhance reader engagement and understanding.

8. Conclusion: The article should end with a strong conclusion that summarizes the main points, answers the questions posed at the beginning, or provides a call to action

9. Call to Action: If the article aims to persuade or motivate the reader to take a specific action, a well-placed call to action is vital.

It’s very straightforward to “ask” AI to help you shape your growing article with these finishing touches.

Let’s see an example from today’s article I am working on for a Revenue Operations blog.

The output was acceptable, and from here I wanted to add density.

I asked for “sub-bullets” and received a little extra.

Now we’re going to take our final step.

Adding a headline, hook, bridge and other finishing touches.

Input:

Response:

We Live In the Age of Ultimate Information Leverage

Not only was the final product lightyears beyond our first nugget of knowledge… the entire process took me 5 real-time minutes.

From here I could get the system to write 5 alternatives of each section, with 3 different tones… and get a beautiful article published.

Of course, you can take this further.

I’ve written a “prompt of prompts” that is basically a program that runs different LLM prompts and then glues the pieces together for me to review.

In seconds.

We truly live in an age of information super leverage.

You can summon knowledge on several domains, build it out, refine it, then fuse these different knowledge domains together along novel seams / lines that you can’t anticipate before starting.

I hope this helps you all gain more from these incredible systems.

The future isn’t in the future anymore.