I have a custom built set of admin tools for APIs.io that help me quickly build APIs.json and OpenAPI indexes for search. While I am a believer in using AI, I am a skeptic of all the hype going on around AI right now. Well trained ML models developed on high quality sources of data and content have the potential to do many useful things. However, I know all too well that when you have garbage into your API models you get garbage out. So, amidst all of this hustle and hype I enjoy having real world uses for applying AI beyond just the images for my stories, and specifically using ChatGPT to help curate high quality API search indexes.
I am processing many API providers and their OpenAPIs. One of the things I do to help enrich the API indexes I am developing for APis.io is clean up the titles, description, and tag each of the APIs, as well as the operations available. The OpenAI API works very well for this. Honestly, it doesn’t take much to write a single sentence of paragraph for each API and its operations, but when you are looking down the barrel at thousands of APIs, it becomes a lot of work quickly. I think this scale and redundancy reflects AI’s sweet spot, not the other areas where we should actually just be doing the work with natural intelligence.
This is the only use for AI I have come across so far. Honestly, my tagging system that parses paths, parameters, and summaries for APIs, then compares that against a vocabulary I am maintaining is much more useful than any AI tagging I have explored. I am sure there will be other uses for AI along this APIs.json driven APIs.io journey I am on. And I am sure that once I have rich APIs.json indexes for all the top APIs out there that training an LLM on this data will be valuable. But not until I (we) have done the work to build a rich machine-readable index of the surface area of the operations surrounding these APIs, and the details of how each of the APIs work. AI will not do this work for us, despite what everyone might think.
Profiling APIs is mundane work. I get why people don’t want to do it. I struggle with it. I also understand the value generated when you do the work. The resulting API and API operational indexes, but also the knowledge I obtain doing the work. I also understand where automation would make sense today, and where it might make sense tomorrow. It is nice to have AI as an augmentation for the work you are already doing. Not a replacement. I feel like if I use it as replace across the board I am outsourcing the value I accumulate to someone else. Writing my API descriptions is about the only use I have for AI today, but is something that will evolve and expand, but at a pace I set, not the industry or markets. In my opinion and experience so far, you are buying into the hype right now you are simply looking to extract value, or you will have value extracted from you, with less than 25% of AI usage being of true balanced value for those involved.
