OpenAI implosion

It depends on what you are going to build. If you are building a simple text generator. Then yes, it will probably not work.  

It is about your SAAS idea. For example, look at my tools here. Let's take the AI-powered keyword research. This has nothing to do with ChatGPT updates because it relies on keyword data.

For me, I don't invest in a project that OpenAI will add as a feature soon. try building something based on the assistant API. or niche-related. I don't know what you are building. but there are different factors that you should study and consider before you go with your idea.

My SaaS ideas revolve around knowledge and project management, along with supervised ML models I train within the Python ecosphere. I am exploring ways I can deploy openai assistants into the backend of a knowledge management SaaS app. (btw, really liked your API course. Appreciate all you hard work and willingness to share.)

Anyway, not sure openai can ever recover (think about all those 100 million and multibillion dollar businesses that are in the process of integrating with openai; they've got to be very nervous) and for sure I do not trust Microsoft, as I've watched the evolution of how they have treated their developers through each "new thing" cycle for the past 20 years or so. Maybe I could hitch my wagon to BART with HuggingFace or some other LLM API. Any suggestions or thoughts? I'm all ears.

I don't know if you have followed my prompt engineering course, I showed there a global LLM generation function that you can use un your apps. Which is integrated wth different APIs and models, so for example, if OpenAI is not available anymore, you will not need to change the app structure, you just connect another API or Model. so It is a dynamic function, so Implement something similar in your code, and you will no be worried about those companies. your app can be adjusted in seconds to work with any model or API.

coming back to your idea, I think it is more complicated than a feature to be added to ChatGPT. if you have your own trained models, data, and so on. I think you are on the right track.

The models I tinker with are supervised learning models, mostly prediction or sentiment analysis. I developed a pretty cool (IMO) api to use google rss news feeds to fetch me articles, which i then parse-to-publish via huggingface pipes.

The KM SaaS is more about backend, agents. Think reference Liberian and other instances I think openai agents could be trained for. 

Thanks for the input. Much appreciated. Happy Thanksgiving, buddy!