Why LLM are failing

Welcome to the metaverse.

Aug 20, 2025, by Ugo

I’ve got a LLM Agent, that reads files and, among other things, extract resources that are in these files. However the extracted resource names are often not corresponding to the files.

Me: Why the resource names are not corresponding to the files ?

AI: The problem is that the LLM is not actually reading the Terraform files correctly to extract the real resource names. It's generating names that sound plausible but don't exist in the actual files.

Me: 🤦


But, but... Aren’t you a LLM also?

← Back to all fails