You are entering a new age of exponential evolution in AI tools. It looks and feels like there are new AI capabilities entering your stack daily - from ChatGPT upgrades to specialized agents that can write your emails, schedule your calendar, and automatically handle routine purchases. The innovations in the realm of pixels seem limitless.
But here's what's really happening beneath the surface: The most striking shift isn't in the technology itself, but in who's using it to compete.
The leading firms in any industry today aren't just dabbling in AI—they're immersed in it. They have deep competence, vast resources, and teams dedicated to leveraging artificial intelligence for competitive advantage.
You are competing against these organizations whether you realize it or not.
The battlefield has shifted. While you were scheduling quarterly planning meetings, solo operators armed with AI workflows began dismantling entire business models. Not through size or funding, but through technical competence and exponential efficiency.
One-person organizations are now outcompeting established leaders. They've built programmatic workflows that replace conventional roles and create assistants that previously required consultant fees. They operate with surgical precision while traditional businesses lumber forward with industrial-age processes.
In this new ecosystem, there are two types of players: predators and prey.
The predators are lean, technical, and relentlessly effective. They build complex, profitable operations that scale without the traditional overhead. Their edge lies in cutting-edge AI implementation and the ability to execute at speeds that conventional organizations can't comprehend.
What they lack is network, long-term relationships, reputation, and founder experience.
This is where you enter the game. Your founder experience—combined with a hunger to evolve your competence—creates a formidable advantage. But only if you're willing to adapt.
My invitation to you is direct: become highly competent with AI or risk irrelevance.
This isn't about merely "staying current." It's about positioning yourself at the cutting edge of a transformation that's reshaping knowledge work potential. By developing these skills, you'll be able to:
Maneuver against or alongside the engineering lions
Maintain your position and influence
See opportunities invisible to those using yesterday's mental models
Execute at speeds that leave competitors wondering what happened
Of all the technical skills you could develop today, prompt engineering stands above the rest. It's the art and science of directing large language models toward specific, valuable outcomes.
Think about it like this: Learning to lead humans allows you to gain influence over them and shape the future together. Prompt engineering does the same, but for artificial intelligence. It makes you the commander of a system that can execute complex tasks at superhuman speed and scale.
In this post, I'll guide you through the first level of prompt engineering expertise: mastering consumer AI applications. This creates the foundation for everything that follows.
Level 1: Mastering Consumer AI Applications
The first level of prompt engineering is where you become a pro user of the various consumer apps out there. Through your usage of these applications, you develop an intuitive feel for how to command these systems to deliver what you want.
What makes this level significant is the quick velocity of mastering these skills through hands-on experience. The natural progression to more sophisticated tools becomes easier and more natural because of your intuitive understanding.
To quickly learn the fundamentals of prompt engineering, there are three core principles to keep front of mind: context, iteration, and model choices. By learning how to command LLM apps through these principles, you create the foundational knowledge to understand at a deeper level how these applications think.
Let's go deeper into the first principle of context.
Context: The Foundation of Quality Output
Adding context to your prompts is probably the most important step when using LLMs to solve any task or problem.
Think about this:
How well would you perform on a task when the description is fewer than 3 lines? Probably poorly. That's because you don't have the required context to create something that transcends the generic. In this scenario, LLMs are equals to humans.
What you must avoid is simple prompts for complex tasks. With every substantial task, you should approach the prompt as a letter to a colleague. Let them know all relevant information, your perspective, and the requirements in greater depth. Let prompt engineering become a practice of your writing skills. The better and more rigorously you can communicate, the greater frequency of quality outputs.
For every project you're working on, there will be context acting as constraints or sources of truth. SpaceX's engineering mantra of "removing first, then optimizing" is a great example of this. When working with LLMs, they can't read the room, and they don't have access to the knowledge or culture code of your team. Now, you have the power to grant access to this information. Just as you lead your team by giving them access to protocols and culture codes, giving the LLMs access to the same information provides them with the coherence of thought to align more deeply with your human team members.
Through prompts and information access, you can install this knowledge as context into the LLMs. The easiest way to do this currently is through projects in tools like Claude Projects or similar platforms. Inside, you can easily upload your team's sources of truth as documents. Then for every message you send through the project, the LLM will iteratively use the documents you have uploaded to understand more deeply you and the context of the project.
You can think of it this way: the LLM has to read your "book of truths" for every message it returns to you. When done right, the results improve the usefulness of the LLM by 10x-100x. Here are three tips to get it right:
1. Simplify: Less is More
Before you throw all of your documents into an AI project, simplify and remove what is not necessary. Removing what is absolutely not needed creates vast performance improvements. In this case, less is more. Leave the LLM with just enough context to work and play with.
What is "just enough"? This depends on the scale of your project (simple or complex). For simple projects, you can add in all the context you want. But if the project is complex and has many moving pieces, then you will benefit greatly from dividing up the entire pool of context into more specific "sub-projects" where the actual context (documents) reflects an isolated moving part of the whole project.
If you divide up a project into smaller projects, make sure you label each project for the specific moving part it reflects. The majority of projects you will work on require only one project.
2. Format: Structure Creates Clarity
Through the process of feeding the LLM your context, you will introduce a vast collection of information and text. You don't want to throw in a mixed salad of information. Creating simple and standard formatting of your documents will enhance the LLM's ability to read and comprehend the context.
The most potent method is to create a good document header structure. This is familiar to the LLM and increases the efficiency of reading, understanding, and then acting upon the context. Ultimately, the time spent on creating quality context will be recouped through less back-and-forth, latency, and confusion.
A good mindset to use is thinking of an LLM as a human assistant whose clear thought and communication is crucial to bridging understanding into assistance.
3. Create Examples: Show, Don't Just Tell
LLMs are example junkies. By providing closely tied examples to what you want, you essentially hack the LLM to do and replicate the example with greater accuracy. For specific tasks, you can add simple examples of “perfect results”. And you can add comprehensive examples to your project knowledge as documents.
Adding examples to the prompts are powerful, but can take time to create. Instead, you can upload a previous “perfect” piece of work for the LLM to comprehend and imitate.
Now the LLM can use these as needed to comprehend the task and in shorter time replicate previous success. For example, if you want the LLM to write in a specific style or format, provide a sample of that style rather than just describing it. If you need a particular type of analysis, show what good analysis looks like in your context.
Real-World Implementation
Let's look at how these principles apply to gaining strategic insights from historical visionaries:
Historical Mentorship
Poor Prompt:
How would Steve Jobs approach my product launch?
Superior Prompt:
I'd like advice on how Steve Jobs would approach our upcoming product launch situation.
Context about our company:
- B2C hardware startup creating a new category of smart home security device
- Currently preparing for our first major product launch after 18 months of development
- We have a working product that outperforms competitors on technical specs
- Our differentiator is exceptional design and simplicity, but price point is 30% higher
- Early user testing shows strong enthusiasm but confusion about certain features
- We have limited marketing budget but strong industry connections
Current launch challenges:
- Deciding between a large, splashy launch event vs. smaller, exclusive preview events
- Messaging is currently focused on technical capabilities rather than user experience
- Considering whether to launch with full feature set or simplified version first
- Debating price point strategy (premium vs. competitive)
- Media outreach strategy not clearly defined
From my understanding of Steve Jobs' approach, he was known for:
- Obsessive focus on user experience over technical specifications
- Creating a sense of exclusivity and anticipation around launches
- Ruthless simplification of both product and messaging
- Strong, opinionated decisions rather than trying to please everyone
- Storytelling that connected products to larger human needs
Please advise on how Steve Jobs might approach our launch strategy given our specific situation, including what he would likely prioritize, eliminate, and how he would craft the narrative around our product.
Historical Mentorship Template
After the Historical Mentorship example above, here's a template you can use to structure similar prompts for any visionary leader whose perspective you want to channel:
I'd like advice on how [VISIONARY NAME] would approach [YOUR SPECIFIC SITUATION].
Context about our company:
- [INDUSTRY/MARKET POSITION]
- [CURRENT STAGE/MILESTONE]
- [KEY PRODUCT/SERVICE DETAILS]
- [MAIN DIFFERENTIATOR]
- [CURRENT PERFORMANCE METRICS]
- [RELEVANT CONSTRAINTS]
Current challenges:
- [CHALLENGE 1]
- [CHALLENGE 2]
- [CHALLENGE 3]
- [CHALLENGE 4]
- [CHALLENGE 5]
From my understanding of [VISIONARY NAME]'s approach, they were known for:
- [KEY PRINCIPLE/APPROACH 1]
- [KEY PRINCIPLE/APPROACH 2]
- [KEY PRINCIPLE/APPROACH 3]
- [KEY PRINCIPLE/APPROACH 4]
- [KEY PRINCIPLE/APPROACH 5]
Please advise on how [VISIONARY NAME] might approach [SPECIFIC ASPECT] given our specific situation, including what they would likely prioritize, eliminate, and how they would [SPECIFIC ACTION YOU'RE INTERESTED IN].
This template allows you to tap into the wisdom of historical figures like Edwin Land, Jeff Bezos, Walt Disney, or any other visionary whose approach you admire. The key is providing specific context about your situation and highlighting the aspects of their philosophy most relevant to your challenge.
Try it out, this is a fun experiment.
The Power of Iteration
Even with excellent context, your first prompt rarely produces perfect results. The real magic happens when you iterate. Think of it as a conversation rather than a one-off interaction.
When you receive a response that's not quite right, provide specific feedback about what's missing or incorrect. Instead of starting over with a completely new prompt, build on what worked in the previous response.
For example:
Initial Response Feedback:
That strategic plan is heading in the right direction, but I need more specific metrics for measuring success in our customer retention initiatives. Also, the partnership section focuses too much on large players—we're actually interested in complementary startups that could help us expand more quickly into adjacent markets.
This approach creates a collaborative dynamic that progressively refines the output toward exactly what you need.
Praise specifically, criticize generally
Model Selection: The Right Tool for the Job
Different AI models have different strengths. Understanding these differences can save you significant time and frustration.
For complex strategic thinking, nuanced writing, and tasks requiring deep reasoning, larger models like o1-pro and Claude 3.7 excel. For quick drafts, straightforward information, or iterative tasks, faster models like Claude 3 Haiku or smaller o1-mini variants often provide better efficiency.
The key is matching the model to the task complexity. Don't use a sledgehammer when a regular hammer will do—but don't try to drive a railroad spike with a tack hammer either.
Moving Beyond Basics
As you master these Level 1 skills, you'll develop an intuitive feel for how to get the most out of consumer AI applications. This creates the foundation for moving to more sophisticated implementations.
The good news is that these fundamental principles—context, iteration, and appropriate tool selection—remain relevant at every level of prompt engineering sophistication. As you grow more advanced, you'll build on this foundation rather than replacing it.
LLMs are getting incredibly smart. But adding your clever use of these principles gives you the power to push them even further.
In my next post, I'll explore Level 2: Building Custom AI Integrations, where we'll dive into creating systems that embody your company's unique knowledge and approach.