Doge Developed Error-Prone AI Tool to "Munch" Veterans Affairs Contracts

Re:Doge Developed Error-Prone AI Tool to "Munch" Veterans Af... by TrapLord_Rhodo on Saturday June 07, 2025 @08:31AM (#44209217)

using an LLM is a very VERY small part of the code he published. 12 lines out of around 13k lines. You seem to be disagreeing with the concept, not the actual code. you prolly could have reviewed the code in the time it took you to google that ben franklin quote.

If you look at the code, he explicitly filters for sole source before he even sicks a LLM on it.

Tell me what you know about far part 6

[ Parent | on HN ]

Re:Doge Developed Error-Prone AI Tool to "Munch" Veterans Af... by ipython on Sunday June 08, 2025 @08:42PM (#44220456)

I'm not sure exactly where you get 12 lines of code from. There are at least a hundred lines of LLM-related "prompt engineering" alone: see https://github.com/slavingia/va/blob/35e3ff1b9e0eb1c8aaaebf3..., for example.

Regardless, number of lines is not an accurate metric of how "essential" a component is to the functioning of the overall system. How about this, try removing the LLM-related code from the script... what functionality is left, exactly?

[ Parent | on HN ]