FounderFight - a fun little rabbithole

// I accidentally published this lol. This is not complete - pls ignore

The Party

If you have read my previous post on rubberduck, you know that there is still a lot that I am trying to figure out right now with product market fit and generally how to make this idea not replacable by ai and be called an AI wrapper (although I don't really think we are). I went to a welcome gathering party in Menlo Park in Silicon Valley and discussed the concept of rubberduck with a couple of people there and it was safe to say they were not as convinced as I hoped they would be and they raised some very important questions that I will need to seriously think about. So, like any other aspiring entrepreneur, I got onto perplexity and started asking questions about the viability of the idea and whether it makes sense. But here's the thing - it's definitely going to be biased towards my idea if it's remotely good, even though it's arguments made logical sense. So I immediately opened up another chat and started asking reasons why it wouldn't work... and it also gave me arguments that made logical sense :) Of course I know there a million factors why a startup works or doesn't work and luck is a very key part but I was just curious on the maximum amount of value I can extract from LLM's on simply evaluating if an idea is any good.

The RabbitHole

I googled, "2 ai's talking to each other", expecting to find a platform where I can put in an idea and get multiple LLM's talking to each other, and also allowing me to enter my own prompts. But to my surprise, there were none ..? It was a very simple thing to build and I'm honestly surprised that there are no ai wrappers that have a paid subscription model that allows more features. I tried doing different keyword searches but it simply didn't exist! Hence, it was in my very nature to then build an ai wrapper that has a paid subscription model that allows more features :) and since I'm quite busy with the move to San Francisco, I employed the help of a good old friend to help me make this. He immediately hoped on board, suggested a couple of improvements (really good ones at that), split the work, assigned a deadline and got to work - agile style.

The Product

It's an AI evaluator that takes a start up idea and gets 2 models to argue the case for and against it. The free version? You get to type in your idea along with one attachment and watch the 2 models fight for 10 rounds! In order to prevent hallucinations / assumptions about your ideas, anytime its unsure about a certain concept or idea, it will pauase and ask you a clarifying question. The premium account will allow you to chose which AI model you would like making each case, have unlimited rounds so you can see the argument become deep and complex, and lastly also provide unlimited attatchments. When you are satisfied with the conversation, there is a judge that can be invoked which will look at the conversation and decide a winner and provide it's reasoning. At all times, you can see its reasoning (or at least the illusion of reasoning). With all this, hopefully people can help evaluate their ideas on a more deeper level, not have a biased opinion from an LLM and maybe even gain new insights! But honestly, it's not that deep. I just wanted to build it for RubberDuck.

The Verdict

Build RubberDuck. // I accidentally published this lol. This is not complete - pls ignore <>