Love the grug-brained demystification here, stripping away all the hype around agentic AI is exactly what's needed. The point about LangChain's SQLDatabaseToolKit being prebuilt really underscores how much of the heavy lifting is already done. What resonates most is your breakdown of the actual complexity, it's not writing the agents but handlig deployment, costs, and governance that separates hobbyists from production-grade systems.
Love the grug-brained demystification here, stripping away all the hype around agentic AI is exactly what's needed. The point about LangChain's SQLDatabaseToolKit being prebuilt really underscores how much of the heavy lifting is already done. What resonates most is your breakdown of the actual complexity, it's not writing the agents but handlig deployment, costs, and governance that separates hobbyists from production-grade systems.
Daniel, you had me at Grug-brained developer. That was one the funniest things I’ve read in a while. Thanks
The most complicated (and important) part, which I think you missed, is how users can trust the LLM's answer (anyone said eval?).
This is a particularly challenging problem - a CFO needs to be 100% confident and doesn't know how to evaluate the underlying SQL.
Thanks Daniel
I’m sharing an n8n agent I also built in one hour including all steps - for free.
https://mariannanakos.substack.com/p/n8n-tutorial-built-an-ai-workflow
I had been staring out to sea, now my wife has me home again.