SEAL + AI, is it possible?

[NiilaM]

As we know, Large Language Models can understand complex code structures and manipulate them as instructed. If we could train an in-house LLM specifically on EBO database structures, code snippets, regional programming guidelines, etc., engineers could simply express their goals in natural language along with reference materials to a purpose-built SEAL agent. The agent would then analyze the EBO database, evaluate what needs to be modified, confirm, and process the changes. This would probably be best suited for working with pre-existing data, modules, or code snippets. 

Such an agent could also provide consultation or insights about a specific system. For example, it could help identify causes of excessive CPU usage, detect binding mistakes, or verify correct configurations.

Overall, this could be a useful addition to the SEAL toolbox, especially for complex, one-off, large-scale modifications or creations that might be challenging to formulate in script. Or maybe not? Is this even possible?

[Benji · Accepted solution]

Thanks! Very good input. While we're not there yet, what you're eluding too is definitely where we're headed.