All I want for Christmas ...
The Gift of Curiosity: Unlocking Subscriber Growth with Simple Questions
You Don’t Have to Be Smart Anymore, You Just Need to Be Dumb Enough to Ask Good Questions!
Before AI, mastering a complex topic meant finding experts to guide you. Now, you don’t need to be a genius to get smart answers. Everything is available a click away. You need the courage to ask questions. But, that’s not all… Let me explain.
I just asked Grok how to grow from 171 (hello!) subscribers to 10,000 by year’s end. It felt like a huge, almost impossible goal. I have an engineering degree and could crunch the numbers myself, but why spend time on that when I’d rather take action and make things happen?
So, here is what I told Grok:
All I want for Christmas is 10,000 subscribers. If I have 171 subscribers today, what rate should I grow at, to get to my desired number by the end of the year? Today is June 3rd, 2025.
Within moments, I received a detailed response that broke down the problem into manageable parts. It calculated the daily growth rate needed and provided alternative strategies to make it happen. The speed and clarity of the answer blew me away. I didn’t need to be an expert; I just needed to ask!
Clever, Maybe too clever?
Grok quickly solved for ‘r’ and came to the conclusion that r=2.755% per day.
It checked its work, and made sure it was correct. It didn’t even stop there, and also provided a number of alternative growth options.
Steady Linear Growth: Consistently adding a fixed number of subscribers daily through regular, high-quality content or targeted outreach.
Exponential Growth: Leveraging viral content campaigns or referral programs to create a snowball effect.
Burst-Based Growth: Timing big wins with seasonal trends or limited-time events, like a holiday giveaway or a collaboration with a popular creator.
Each method came with actionable ideas—think viral video challenges, tapping into community-driven momentum, or riding the wave of a trending topic. It was like a buffet of options, all tailored to different strengths and resources. Each post was a chance to connect, inspire, and grow.
I was set to be the number one newsletter on Substack. Wonderful!
But was it, wonderful?
Do you see anything wrong here?
Grok, unknowingly to itself, made a mistake in its calculations.
1.02755^211 is 309, not 58. The actual growth rate should be about 1.95%.
If you’ve been playing with LLMs long enough, then you know that it’s a well know fact — LLMs are very bad at math. They don’t actually DO math, they just predict the answer, which for anything more complicated than 2+2 will obviously fail. The way to get around this is to have a multimodal LLM that can spin out a code environment, and do the calculations in a verifiable, unmistakable way.
Usually, all the modern iterations of AIs do just that. Grok, OpenAI, Anthropic at el, all the “thinking” models know to do math by running a dev environment, but in case, something failed. It’s an easy fix though. I pressed Grok, and I asked it to write a Python code to check its work, and to correct itself. What really surprised me — Grok pushed back!
It wrote the code, and still proceeded to tell me that the wrong answer was right. Why? It did not actually evaluate the code that it wrote. It just continued with the assumption that the previous answer was correct.
If only it knew to hit “preview” on its own code, it would’ve known that it made a mistake, but it did not.
The Lesson: Trust, but Verify.
As it turns out, while you can learn anything using LLMs, we aren’t yet at a place where you can blindly trust all the answers.
Asking questions is only half the battle.
You also need to verify the answers, especially when math or complex logic is involved. LLMs like Grok are powerful, but they’re not infallible. They can point you in the right direction, but it’s up to you to double-check the map.
If you know the tools to check the answers, use them. If you don’t know, the simplest way to check one LLM is to use another LLM to solve the same question. Sometimes, it works. Unfortunately both free OpenAI model and LLAMA 4 face-planted on the same question when I asked them.
Sometimes, a good old calculator (or some code) are still the best. Just plug in some numbers, and see which ones yield the expect result. Yet again, you don’t need to be very smart, you just need to pay attention. Eventually, the LLMs will get good enough, we are just not there yet.
Note: If you want to be a little geeky, you can use WolframAlpha to plot the function and get a visual result to check your work.
What’s Next?
I’m now armed with a corrected growth rate and a handful of strategies to hit my 10,000-subscriber goal. Whether it’s posting consistent content, launching a viral campaign, or timing a big event, every step counts. Realistically, I don’t think I am going to hit the near 2% daily growth. But a boy can dream!
Meanwhile, I’ll keep asking Grok for help along the way. Even with occasional mistakes, I move so much faster when I have AI at my fingertips. After all, if I stop playing with AI, what are we going to talk about?
Good night,
Kirill.
What’s your next question? The possibilities are endless. Ask away, check the answers, and watch your goals come within reach.