Strategy and the LLM Bot

I stumbled across an article last week discussing the use of vibe doding to generate solutions for specific niche businesses.  Sadly, I forgot to save the article, because, while I wasn't sold on the vibe coding aspect of article, there were links to another set of articles that helped identify boring niches.   I was somewhat surprised by the example in the article, because the author wrote something for time tracking.  I mean, I know that, for some reason, all time tracking software solutions everywhere inexplicably suck, but it's not really an undiscovered problem.  I was expecting to see something far more obscure, like a site that produces documentation and invoicing specifically for small companies that recycle used cooking oil.  Now there's a niche you can exploit!  

This spurred me to query Gemini for "software for boring niche business ideas."  I'm still pondering delving into that more, but a lot of the advice was - predictably - fairly generic.  One of the suggestions was invoicing and time-tracking for dog walkers.  Call me skeptical (and admittedly, I haven't done the research), but I'm guessing someone has likely carved out this niche. 

[Oops, nvm, now I did do the research - turns out there are several options.  I looked at one of the top links, Time To Pet, and noticed it's a subsidiary of a larger company.  The site charges between $25-$75/month per user.  Seems like a pretty lucrative niche, especially for a small business.  With only 400 users at the lowest tier, that's still $120K a year.  I think most people would be happy capturing $120K in revenue per year (or $10K MRR as the LinkedIn kids like to say).]

As nice as it'd be to have an LLM generate niches for me to investigate, I was really looking more for strategies that would help me find niches.  I tried a few follow-up queries, but the LLM was still attempting to overhelp me by telling me what niche to select vs. how to select it.  Remember, these are machines that can only regurgitate combinations of text they've been trained on, so chances of encountering a novel solution here are slim.

After some generic back and forth, I decided to test out its reaction to something more specific - Exhibit Meh.  I told it something along these lines:

I'm creating a mini art history site that contains a few interesting facts per piece.  I'd like to spend no more than 20 hours per week on its upkeep and am looking to make between $15K and $30K per year out of the operation.

Surprisingly, the Gemini 2.5 Pro version gave me better suggestions than I've seen from any LLM, and I'll even use some of those suggestions for attempting to build a more robust business in the coming weeks.  Unsurprisingly, Gemini 2.5 Pro only allows a certain number of tokens for free before gating the user until the following day to continue the conversation.  Though it comes with a few additional Google-themed perks, I'm still not at the point where I'm willing to shell out $19.99/month for an LLM. 

However, it did prove fruitful in this session.  The idea that's got me most excited is a subscription model with premium content.  I've been stewing over how to provide something premium without degrading the existing experience for users.  Gemini suggested targeting 400 people at $50/year to subscribe to a premium newsletter.

I'm still not confident my content is worth $50/year, but $25 seems doable.   Eventually, I'd aim for a user base of 400 to 1200 people.  Gemini quite condescendingly warned me away from numbers greater than 1000, claiming that the difference in its recommended 400 users and my desired 1200 person base amounted to "non-linear" growth.  How it can make this assessment based on 2 data points is beyond me, but it did, and used a metaphor of cooking for a large party to emphasize its point.

I can't say that I'll dodge all of the pitfalls associated with potential increases in a user base, but I'm hoping that my experience leading a team that could accommodate 25K hotel searches per minute means I have at least a passing understanding of scalability.  

Now, the LLM pointed out that the scaling problems weren't likely to be technical, but customer-based, which is a fair assessment from our little steel-hearted tech-bro.  But given the offering I'm presenting on the platform that I'm using, I sincerely hope that I won't be overwhelmed by customer support calls.  I know that things go awry, and I know that some people are just insufferable regardless of how well they're treated, but I'm going to guess that 1200 people who willingly sign up for my particular perspective aren't going to be jamming the switchboards with demands.  If I reach issues with scalability at that level, I'll deal with it, but I have to (a) create premium content and (b) get my first customer before I concern myself with my 1200th.

As usual, I've buried the lede, and you're eager to ask - what is this premium content?  

To start with, it'll be an accompanying blog post (very similar in style to, say, this one) that covers additional facts I uncovered while researching a piece, and additional meandering thoughts that are relevant or pseudo-relevant to my discovery process.  Gemini had suggested giving away the "first blurb on the piece" for free and restricting others, but I don't like the idea of restricting content that's currently available to everyone.  The benefit of a premium membership should come with additional content.  

I expect I'll come up with other content ideas, but that one's pretty solid as a starting proposition.  It doesn't make me feel icky by bilking potential customers, and it at least has a puncher's chance of succeeding.  It requires work, but not something I can't handle on a weekly basis.

Plus, if it succeeds at all, it'll be a great learning exercise that will allow me to document the trials and travails of launching a small website with greater confidence.

Good job, Gemini.  It only took an LLM 17 months to simulate something novel that I can use to propel my business forward.

Until next time, my human and robot friends.