Jason Cohen says there are currently three categories of AI products:
- Existing product enhancement
- Expert enablement
- Muggles empowerment (the new no-code)
He argues, and I agree, that the only category that AI is ready for is the middle one. The first doesn’t work, or at least doesn’t add dramatic value. The last only gets the user 80% of the way to a production product and then they are stuck.
Expert enablement means that the person who can do the thing does 80% 100x faster with the help of AI and has the expertise the finish it with or without the help of AI. This is how AI should be used in Graffiticode. Whereas we make a platform and languages for enabling domain experts to solve problems with code, we now have the ability to train AIs to do that for those experts. And, because they are experts, they have the ability to either write the code or coax the AI to finish it correctly.
This works because Gc provides all the hard bits developed by software experts who can build the language and foundation correctly so that the domain expert doesn’t have to, whether they are typing code directly or through the AI.