My first hackathon at South Park Commons x OpenAI
I participated in the Open AI x South Park Commons hackathon two weekends ago. It was my first hackathon and I loved the energy and how it felt like it was possible to build anything.
These were three things I took out from the experience:
The barrier to entry to incorporate GPT and LLMs is extremely low.
User experience differentiates your product.
Side projects are a fuel for creativity.
The barrier to entry to incorporate GPT and LLMs is extremely low
As a non-engineer, I didn’t realize how easy it was to incorporate AI into your product. There is virtually no moat.
OpenAI has made it extremely easy for developers. You just need to create an OpenAI account, set up a virtual environment, create an API key and then send in your first API request. It’s really that simple.
The hardest part about a hackathon is that everyone is committing code at the same time. There wasn’t any need to build anything revolutionary, just use a lot of existing libraries and components and then set up relevant API endpoints.
Rather than AI differentiation, there were two other criteria to consider as we used AI to build a product:
Costs. The importance comes into how you use the GPT and then also being super cost conscious for any token limits for user inputs or outputs. If you’re not careful, it can be incredibly expensive and drain your OpenAI account.
Distribution. Because a company’s moat is no longer from the tech, it has to come from something else. NFX published a recent article comparing AI to the bottled water industry (reference). AI-based companies differentiate themselves from their distribution, rather than tech. If I’m targeting enterprise customers, I win in the market by having a unique entry point with customers and a clear conviction on the problem that need to be solved. If I’m targeting SMBs, my marketing approach might differentiate me from other 23 competitors building the same thing.
User experience. User experience actually does continue to be important to stand out in the competitive landscape, which brings me to…
User experience differentiates your product
Reading, Justin’s Technically article this week made this click for me. I found his analysis between OpenAI and Mistral intriguing for how user interfaces are critical for better engagement and customer delight, though people don’t talk about it as much. I would recommend reading his article but the comparison between GPT4’s user interface versus Mistral’s was helpful.
Generally, on Chat GPT, users have the ability to:
See previous search results
Ability to easily upload a file/document
Picker to switch between models
I would say things they could do even better as someone who is a power user and needs to organize my workspace:
Folders for various search results, especially as ChatGPT is used as a prosumer product and I want to separate how ChatGPT is assisting me for work versus various side projects
ChatGPT 4 counter: I want to know how close I am to hitting the maximum messages that I can use as a Pro subscriber out of the 40 messages, and then a time countdown when I’ve gone over the number of messages sent to ChatGPT
On Mistral, the user asked to create a tweet thread based on a blog post link. Instead of showing an error state saying it couldn’t browse the internet, the model started hallucinating because it couldn’t browse a link . There generally were error states or dead ends that it couldn’t help the user navigate around (though with rapid iteration, they might fix it).
In a nutshell, a user interface is a strong way of building user retention and engaging with customers effectively. UI hasn’t evolved yet much from how we interact with current products or Apple interface guidelines and Google design guidelines but I imagine that is where OpenAI and other foundational model companies are evolving UI to differentiate themselves.
Side projects are a fuel for creativity.
This was my first hackathon and I would gladly do more!
As a non-technical person, I was quite scared about participating in my first non-company sponsored hackathon. But it was really worth it. I loved committing code (though to be safe, I limited my blast radius to simple changes).
It’s also a lot easier to build MVPs with no-code or low code tools. Our team relied on Streamlit to build out the initial product. If you’re interested in the product, I have teammates who are continuing to build out the product! Check it out.
https://sherlockhomie.com/
Kudos to the team that won Grand Prize for the hackathon. Memeverse translates webpages into video meme explainers. Take a look at the demo here!
How have you been experimenting with AI?