About AI Report Library
In short
- There are too many AI models and settings, and they change too fast to test them all yourself.
- This site shows how the research outputs differ by model and other variables, for non-coders doing market, buyer and customer research.
- It’s for solopreneurs and small business owners who do it all themselves: sales, research, marketing, regulations, new business ideas, etc.
- Don’t fully trust what AI spits out. I don’t either. Use this as a reference.
The reason why I created this: when I try to use AI models to do some research, there are too many options right now. Models, reasoning levels, projects, functions, you name it. It’s very hard to find the right one, so I ended up picking the two major frontier models, ChatGPT and Claude. The models are also changing so fast that testing all of them every time a new one is released takes too much time and a lot of work, aka time consuming and token consuming.
There are a lot of benchmarks for coders, but for non-coders, including myself, the work is more about qualitative tasks: understanding the market landscape, understanding buyers, customers and targets.
This comparison website is for solopreneurs, and also for small businesses where the CEO needs to be the salesperson to sell the product or services.
Hiring assistants is still a lot of investment for you, but you can spend a couple hundred dollars on LLMs so that you can maximize your solo capacity and boost your sales and research capability as much as possible in this AI era.
While I’m using AI for a large part of my business, I don’t really fully trust the output that all these LLMs spit out all the time, and I check the sources myself.
There are of course hallucinations and mistakes that AI makes, outputs with low accuracy, and also the speed that is slow sometimes. And when the models are updated, there are crashes and some malfunctions right after the new model releases.
And also there are a lot of fake AI influencers around social media, and you can’t even figure out which information is correct and which is not.
One of the reasons why there is nothing like this for qualitative information is that web search is unquantifiable. This is what I believe in the first place. When you quantify the unquantifiable, it becomes subjective. I understand that. Yet sharing the output differences by model and other variables will give you a rough picture of what those combinations spit out, simply and mainly as a reference.
So that’s it. I use these tools every day for my own business, and I’d rather spend the time and tokens once and share what I find than have everyone test the same things over and over. Take what’s useful, check the sources yourself, and spend the time you save on selling.