ChatGPT Open Access Wrapper
Publicly building HORAI: a ChatGPT Wrapper to improve customer experience
When I started to learn more about the ChatGPT API I wondered why nobody is building a white label wrapper yet? It was the first idea that came to my mind. Maybe it’s stupid? My thinking is: Companies want to be part of the paradigm shift that AI is injecting into our daily life. They want to appear future proof. They even want to have their own ChatGPT, to enable their customers to use it freely? As of today I am not sure if that is really something that companies would buy. To be honest: I don’t care. Building things is fun and I learn a lot. It’s win-win for me, even when it is not making any money. So here is me, building HORAI: the first horizontal AI!
The value
It is all over the internet. Even on TV, people learn about ChatGPT and what is capable of. But there is a wall. It’s a rather small one but a wall nevertheless. You need to register/login to use ChatGPT. Take that away and put a logo of a company on it. Now a company suddenly has a new service layer: an AI chat with your name on it! The company can provide some nice predefined (company focused) prompts to kickstart the conversation.
The name HORAI stems from the idea of using a horizontal AI. The opposite of a vertical AI (which would be trained on the deep (not openly available information) level knowledge of a company). ChatGPT uses publicly available information and wraps that up. No need for a visit of the companies website and subpages anymore. The beauty is have that as a possible front page. It’s like the door holder and greeter in the physical world. Only he knows at which isle you’ll find the specific cable you’re looking for.
The status
I have built two MVPs. One with Bubble (no-code platform) and one with a react app boilerplate. Both working as intended. The react app is faster on the client site and uses Tailwind CSS, which I definitely prefer. Also when it comes to scalability and pricing. Bubble is expensive. Having a react app would kind of kill two birds with one stone. Now I am not a developer, therefore it takes a lot of time to figure stuff out. I like the learning curve but I feel, having someone who knows all the yaddayadda of front/back-end engineering makes sense. I’d rather focus on other things. Like the UI and the marketing. In these disciplines there are different kind of challenges: They currently look very similar to ChatGPT what I definitely will change. The idea is to make it more accessible and that includes several future steps.
Feature ideas
Many ideas pop up while working on the project. A headless CMS for managing the pre-defined prompts. A user-customizable white label UI. An API-Key manager UI. Captcha/IP-limiting/etc
Thinking out loud
When I studied Interfacedesign on Potsdam there was always this one prevailing opinion: the internet as a human readable information layer is only a minor transitioning phase. Web pages are holding information and interactions since 30+ years. We are used to input an URL or follow a link, and that is basically how we access and organize our collective knowledge. As an Interfacedesigner it is kind of frustrated to see in what we have locked ourselves in. Humans are able to communicate on so many different layers. A web page is not human centered. It is text and images on a canvas. Like Teletext on a TV screen. We can do better.
With ChatGPT we now have a different way of asking and retrieving information. We interacting with an entity rather than clicking buttons and links. The entity has a more human friendly approach: the conversation.
TBC