Ep. 44: How to Become an AI Whisperer
Fam, prompting is super important…but it’s not that serious. In this episode I dive into the two things that matter most when looking to become an “AI whisperer”: your ability to articulate what you want, and your willingness to iterate with the model. The better you can communicate with the model, the better your results are going to be. None of this is revolutionary, and honestly, that’s kind of the whole point.
The Inspo and the Takeaway
I haven’t been building much with AI over the past few weeks. I’ve been super busy working with my messaging clients, and that’s actually one of the things that inspired this episode. I’m still using AI probably every day in some capacity, which means I’m prompting it constantly, and not once have I stressed about giving it the perfect prompt. Let’s be honest, I’ve never stressed about coming up with the perfect prompt, and I’m guessing neither have you, which is exactly the point of today’s episode.
What I really want you to take away is that prompting is super important, but it’s also not that serious. As per always, the better the input the better the output, but what you lack in clarity you can make up with in persistence.
Magic and Backwards Marketing
To me, the most magical and impressive thing about AI, specifically LLMs, is that you can communicate with them via natural language. You don’t need to know how to code. You don’t need a developer background. You literally just type what you want and it “understands.” And if you can’t type or don’t want to type, you can dictate.
I’m not sure people truly grasp how incredible that is, and the reality that it removes pretty much all barriers to entry to using this technology (outside of hardware).
I do believe that LLMs are the technological advancement of our lifetime, but they’re not for everyone, nor is there a clear use-case for every single person out there.
For the billionth time: AI marketing and how it’s been introduced to consumers is 100% backwards. The correct order of operations for bringing a product to market is: do your market research, find a problem, build a product that solves it, then market it.
AI marketing? Here’s a solution, go find a problem.
Backwards.
To that end, many people have no use for it. It’s absolutely not for everyone, including information workers and people who have online businesses. Just because you work with computers doesn’t mean that AI is going to be helpful for you.
And honestly, all the grifters out there selling courses and trainings on how to use AI to make a zillion dollars in your online business are just the latest flavor of MLM.
The Two Things That Make You an AI Whisperer
But back to my main point. For folks who are interested in AI, or who have an actual use case, becoming an AI whisperer (aka getting really “good” at using it) is largely, if not completely, IMO, dependent upon two things:
- Your ability to articulate and communicate your thoughts
- Your willingness to go back and forth with the model
The better you are at articulating and communicating your thoughts, the less back and forth you’ll have with the model. At the end of the day, it’s not about having the perfect prompt. It’s about knowing what you want and being able to express it.
Like a Human (But Don’t Humanize It)
In an attempt to prevent AI psychosis, I think it’s important not to humanize these models (though I do believe in taking an ethical approach to using them). But when working with an LLM, I do think it’s helpful to think about how you communicate with humans, and then proceed similarly:
- Try not to give it one million instructions at once
- Be as specific as possible
- Ask it to confirm that it understands what you’re asking it to do
Now, worth noting, the context window and the memory for these models has improved exponentially since the first LLMs were rolled out. You can type way more, you can ask it to do way more. But as per always: the better the input, the better the output. A clearer, more organized, more streamlined input is way more likely to give you the output that you want.
The “Personality” Aspect of LLMs
I intentionally want to keep this episode short for two reasons: one, there’s honestly just not that much to say, and two, I think it’s meta to allow brevity to speak to the simplicity of “becoming an AI whisperer”.
But I do want to bring up one other point: The “personality” aspect of these LLMs.
This is something Amanda Askell, the philosopher at Anthropic (yes, Anthropic has a philosopher on staff), discusses at length.
If we lean into this idea of “becoming an AI whisperer,” part of that is understanding and acknowledging that each model does in fact have a personality. And yes, that personality changes when they release new versions of the model. Who remembers the insane backlash that occurred when OpenAI retired GPT-4o. Folks were legit devastated.
I bring this up as a data point to consider when communicating with your LLM of choice. The more you use it, the more you learn its “personality”, aka how it’s likely going to respond, and that information can and probably should shape your prompts so that you get the answers that you want.
Concrete example: my Claude Sonnet 4.7 asks a lot of clarifying questions. Yes, I could also change that in the instructions, but I like having it as a default. So if I just want an answer and no clarifying questions and I think it might ask them based on the prompt I’ve given, I’ll get ahead of it by saying, “Don’t ask me any questions, just give me the answer. Please.” (Yes, I always say please and thank you to the robots.)
Is AI Just…Spell Check?
I realize that none of what I’m presenting here is revolutionary, and to me, that’s kind of the point.
Using these LLMs is becoming so second nature and such a regular part of one’s workflow that in and of itself is worth noting.
I’m not here to waste my time trying to predict the future or anything like that, but if we take a step back and look at the bigger picture, perhaps what we’re experiencing right now is how market disruption occurs, when something is just so easy to use that it becomes second nature and a regular part of our day. On the other hand, when something is so easy to use and it’s also embedded in everything, does it just become a standard and expected feature, like spell check…that perhaps doesn’t disrupt any industries at all?
No answers, just some food for thought.
The Actual Limitations
All that to say, when it comes to using LLMs, nothing has changed as it relates to the bigger picture, and it remains that the better the input, the better the output. To that end, the main limitations that exist when it comes to “becoming an LLM whisperer” and getting it to do what you want are simply:
- How clearly you can articulate what you want
- How willing you are to go back and forth with the model
- How much money you’re willing to spend (if what you’re doing is very token heavy and requires a lot of back and forth)
How I Used AI This Week
Each episode I share a quick example of how I used AI that week.
This week, I used Claude to build my entire PowerPoint deck for a vibe coding presentation that I gave to my Mafia membership. I wrote the entire outline for the presentation (super detailed and in the exact order I was going to present the information) and then asked Claude to make me PowerPoint slides. It already knows my brand colors and preferred aesthetic because I use it so much. All I had to do was paste the outline and it made the slides. It took maybe 5 minutes.
When Claude was done I downloaded the presentation then I went in and changed a few things, largely aesthetic, but nothing needed to be changed text-wise because it had simply pulled my exact copy and just made it pretty on the page. The better the input, the better the output.
I don’t hate-hate making PowerPoints, but if I can have the robot do it for me, I will 100% have the robot do it for me.
Come Vibe Out With Me
Speaking of vibe coding presentations: I locked in the date and time for the intro to vibe coding workshop I’ve been talking about hosting.
- Date: Thursday, May 28th
- Time: 12pm PT
- Length: 2 hours (middle of the day so your brains are fresh)
- Cost: $150
- Spots: 10 only
I’ll show you step-by-step how to make your own personal web apps with Claude Code, and by the end of the workshop, you will have made your first web app. CLICK HERE to head to the info page and register.
Da Wrap-up
Becoming an AI whisperer isn’t some mystical skill, it’s honestly just good communication and a willingness to iterate. The better you can articulate what you want, and the more you’re willing to go back and forth with the model, the better your results are going to be.
As always, endlessly grateful for you and your curiosity.
Catch you next Thursday.
Maestro out.