What do you make of AI API wrapper products?

Muhammad Roushan
5 replies
We've seen a massive rise in apps that do a lot, but essentially, do very little on their own. A lot of them are reliant on OpenAI APIs - or any other public API from Replicate. If you're someone who is developing one of these products or is currently running them - what concerns / fears do you have? Do you look at it as a short-term opportunity until OpenAI drops the other shoe? I'm really curious about the longevity of these products, and even more curious about the vision of their makers - let's discuss!

Replies

Takeshi [ @Bootstraptor ]
On one hand, this approach allows for a swift and cost-effective start, which is particularly crucial during the MVP validation stage. In case the project exhibits growth, there is always the option to lease a dedicated GPU. Let's take my side beta project, https://Remixo.ai, as an example. With this setup, I don't have to pay for GPU usage when I'm not actively rendering. Instead, I only incur costs for the rendering time itself, rather than the total duration. On the other hand, if the API server experiences downtime, which can occur even with Replicate at times, it can create a sense of being trapped, especially when you already have paying customers.
Muhammad Roushan
@bootstraptor Good product, first of all. I agree - when it comes to AI API apps based on models that are also available on HuggingFace and GitHub - there is always a possibility of deploying them on your own machine etc. But with apps that are solely based on the GPT-4 API, something that is exclusive and has no open-source alternative, you'd feel really trapped if they were to simply take it away at any point. I guess my train of thought here is - you don't have control over it. Some products are very much dependent on a single API that can simply go offline tomorrow. For someone like me - that's a HUGE red flag..
Takeshi [ @Bootstraptor ]
@thisisroushan that's why it's so important for the community to develop open source GPT alternatives (that already exist), but for the time being, of course, they are inferior to corporations gpt's
Austin Nguyen | Afforai
Let to the convo but I think that the app should try to do something different from the base OpenAI model. I try to move away from making a quick cash grab but giving the user some sort of value. Take for example what we are doing rightnow Afforai. What we are doing differently is serving the market in developing countries - starting with Vietnam, with an enhanced model from OpenAi which can read documents in English but be asked about in any language, effectively removing the language barrier for user.