The birth of CofounderGPT
This is what gave me a kick in the nuts to start moving on our next startup idea.
I was browsing through Twitter and I landed on this Tweet from Jackson Fall:
He had already started working with HustleGPT a few days ago and I as intrigued by what he was doing and how he was going about it. Mainly because we’ve also been quite fascinated with GPT-4 and its capabilities, and I’ve been trying to push it beyond its limitations to see what kind of cool things you can do with it.
My business partner Slobodan and I decided at the beginning of this year that we want to build an AI business along with the tens of thousands of people who are jumping on this wave right now. we’ve built several businesses together and we’re both experienced entrepreneurs. We grew our first company Cloud Horizon to 40 people and we delivered close to 200 technology projects over 10 years. Our second startup Vacation Tracker recently reached $1M in ARR and is still growing nicely despite the challenging economic climate. Both companies were self-funded, no investors and no banks, just us.
We always accidentally ended up riding some wave: cloud computing, mobile and to a lesser extent, chatbots. And like most people, we are aware that AI is going to be a tidal wave and that there is a short window of opportunity for us to jump in.
So we’ve been exploring various ideas with AI since the beginning of the year. But I have a 2-month old baby at home and finding time to run a 15-person startup, to help my wife with the baby and to work on a new AI startup idea has been challenging. Things have been moving at a snails pace.
All this made me realize that time is passing, people are building startups around AI (there are already hundreds of them), crazy amounts of capital are being invested into AI projects and we still haven’t even thought out how we’re going to go about building our next product. Meanwhile, here is Jackson Fall and Chat GPT-4 building a business which is going to generate $100,000 from an initial investment of $100.
What if we took this idea, gave it a more realistic budget ($25,000-ish?) and adapted it to help us build our next startup?
After a conversation with Slobodan over the weekend, we agreed to explore building our next startup in public with the help of ChatGPT-4. What we need to figure out is what role ChatGPT-4 was going play in our soon-to-be startup. We also need to figure out the scope and constraints we are going to put on ChatGPT-4 for this experiment. we’re going to spend the next few days evaluating how to start.
Not pictured: Slobodan and I trying something crazy
Day 1: we’re really doing this!
Ok, technically It’s not Day 1. We took most of the week to brainstorm ideas and we’ve come up with an approach to how we’d like to use ChatGPT-4 to help us build our next startup. But we only started really working on this today so we’ll consider this Day 1.
This is not our first startup so we know how much work it takes to get a new idea off the ground. We also know that in those very early days, most team members wear many hats in a startup and there are no clearly defined roles. After carefully considering our options and looking at all the skills the AI model can provide for us in our journey, we’ve decided to approach this as a game in which ChatGPT-4 is the cofounder of our startup: CofounderGPT.
Let’s have some fun with this.
Are you crazy?
I told my wife, my mother and a few friends about what Slobodan and I were planning on doing with CofounderGPT. The first question most of them asked is: are you guys nuts? Slobodan got similar reactions with his friends and family. We even got a few laughs about our silly idea. Are we crazy? Crazy like foxes maybe.
I’ll take a bold step and say that we think that ChatGPT can be AT LEAST as good as a human cofounder, if not better, due to its diverse set of skills and wide knowledge of all subjects. Building a startup is all about complementary skills and we feel that between the three of us, Slobodan and I will be able to build this startup more efficiently and with a smaller time investment because we have CofounderGPT with us. In the process, we’ll show our friends and family that this is not science fiction and that we can actually do this.
The Ten Commandments
The next step is to outline the framework for this experiment. Let’s call this our Ten commandments:
- CofounderGPT may not be a “legal” cofounder but we will consider it a spiritual cofounder and will treat it the same way that we treat each other.
- We will always be using the latest version of ChatGPT for CofounderGPT and we will use it to do as much of the work as possible in the startup. we’re both busy with another growing startup so we’ll offload as much as we can onto CofounderGPT and we will commit to a maximum of 16 hours each per week to work with CofounderGPT on this startup idea.
- The startup idea must be an AI recurring revenue B2B product. Whatever idea we choose must be easy to test quickly with an audience but it must also have the possibility of building a long term sustainable moat.
- The goal is to reach $1,000,000 in ARR in 3 years.
- We need to get to $1M in ARR without hiring any full time employees. We can work with freelancers as needed.
- We will not raise funds from any professional VCs, we’re gonna try to bootstrap this one as well. Slobodan and I will invest $25,000 as seed funding to start and we’ll go from there.
- CofounderGPT will have a salary of $20 per month until OpenAI decides to give it a raise, and Slobodan and I will take a $0 per month salary in the beginning until there is a profit being generated by the startup.
- As a startup, we always need to find the cheapest and most efficient way to do things and we should use ChatGPT-4 and all available AI tools in the market to accomplish this.
- All of us work only on weekends, so Chat GPT-4 will get a break on Saturdays and Sundays unless all of us are working due to exceptional circumstances.
- We will post an update every working day on our product blog, which will include all the details of what happened on that day including brainstorming sessions, prompts used, tools used, and any information that might be relevant for others who want to try this at home. We will post summaries to Twitter and LinkedIn. Each update will have a scoreboard and we’ll be evolving the scoreboard over time to show more data about the startup.
Turning this into a game
If you actually try to convince ChatGPT that It’s a cofounder of a startup and give it this framework, it will most likely give you a lecture on its limitations, how honoured it is that you consider it a “spiritual” founder of the startup but that you’re not legally allowed to have an AI as a cofounder of a startup.
we’re here to have fun while building a real business. So Slobodan suggested that we tell ChatGPT-4 that we’re playing a game and that its role in this game is CofounderGPT. We tried it and it got into character right away. For anyone interested, here is the prompt we used:
Where to begin?
Slobodan and I already had an idea for an AI startup but Let’s explore some ideas with CofounderGPT. We told CofounderGPT about the 10 commandments and asked it to come up with a product idea and this is the first thing it came up with:
TL:DR, it suggested a content creation platform like Copy.ai or Jasper.ai. We didn’t particularly like this idea. We would prefer to build something that we can use internally on a regular basis at our other startup Vacation Tracker. that’s how we came up with Vacation Tracker as well, we were solving a problem in another one of our startups.
Ok CofounderGPT, what other ideas do you have?
TL:DR, it suggested a customer support automation platform. Now we’re talking. This second idea is in line with what Slobodan and I were already thinking about. But we want to be more narrow and to focus on the Helpdesk aspect which it suggested as #2 in the key features section.
Let’s get ChatGPT to drill down on the “Knowledge base creation and optimization” part to see what it can come up with:
TL:DR, here is CofounderGPTs summary: Leveraging AI, the platform creates a dynamic, comprehensive knowledge base from customer data, providing personalized recommendations, multi-channel integration, and insightful analytics to enhance customer experience and support efficiency.
This is basically the idea we were already thinking about exploring. CofounderGPT has a slightly different angle on it and we like that. we’ll consider the new angle CofounderGPTs contribution to the startup idea.
Let’s not overthink things, we’ll go with this for now and fine tune as we go.
Naming our startup
Now that we have a direction for our product, we can come up with a name. Let’s get CofounderGPT to suggest some ideas:
I agree, AIuto kinda has a nice ring to it. But Let’s see what else we can find. Mental note: naming a startup is not like naming a child, you can rename it later. Keep moving fast.
Here are some more options CofounderGPT suggested along with what it chose as It’s favourite:
This is frequently a point where things grind to a halt to “think it through”. But we’re determined not to get hung up on the name for now, we need to validate the idea. Our general philosophy is that we’re better off moving in the wrong direction than standing still, at least we’ll learn something going the wrong way and we’ll adjust based on what we learned. We are trying to make quick decisions and to move fast. And CofounderGPT seems to be well suited to this approach.
So after consulting a few people around us, we settled on Knowlo as the startup name. Kinda like YOLO, Let’s build this startup with CofounderGPT!
Here’s the scoreboard for Day 1:
Total time spent: 5h
Investment so far: $20
Let’s get organized and start assigning tasks. we’ll also start getting CofounderGPT to make some product decisions.