Hacker News new | past | comments | ask | show | jobs | submit login

>> RAG pipelines with your data

FTA

"Everyone is talking about Retrieval Augmented Generation, but most companies don't actually have any internal documentation worth retrieving. Fix. Your. Shit."

> 5. Code generation /Review

FTA

"If another stupid motherfucker asks me to try and implement LLM-based code review to "raise standards" instead of actually teaching people a shred of discipline, I am going to study enough judo to throw them into the goddamn sun.

I cannot emphasize this enough. You either need to be on the absolute cutting-edge and producing novel research, or you should be doing exactly what you were doing five years ago with minor concessions to incorporating LLMs. Anything in the middle ground does not make any sense unless you actually work in the rare field where your industry is being totally disrupted right now."

The man trots out his bonified's at the start and in the article. He's inside and backing up that rage.




"Everyone is talking about Retrieval Augmented Generation, but most companies don't actually have any internal documentation worth retrieving. Fix. Your. Shit."

I read up about Copilot as part of some internal research, and absolutely the first things to do for Copilot (I'll copy-paste just the line-item headings from the section entitled "Prepare your data for Copilot for M365 searches") is:

  - Clean out redundant, outdated, and trivial (ROT) content.   
  - Organize content into logical folders and sites.
  - Tag files with keywords.
  - Standardize file names.
  - Consolidate multiple versions.
  - Promote data hygiene habits.
Sigh. If I could do all this at an organizational level, I wouldn't need copilot at all.


What the hell does that mean fix your docs?? -

companies have tons of document libraries and documentation that need sifting through and are generating more content regularly and => RAG and vector search is game changer with real value there

Eg we implemented RAG + vector search at a manufacturing company and it changed their workflows entirely

And to coding with LLMs, Say what you will about AI coding but code review/linting and LLM created unit tests are as itself game changing as IDE intellisense this value is worth at least one junior developer on the team - that’s 70k yearly salary alone benefit


> What the hell does that mean fix your docs??

If you asked a company to run for a week, based on what its docs said, and nothing else I suspect it would be bankrupt before Friday.

Knowlege is tribal, human, and adaptive.

AI did this once already. Harvesting the data from professionals for expert systems... The problem is you need to keep feeding it data... the people dont go away, they aren't doing the job any more, they are just documenting the job at that point.


A measly 70k saved today to make fewer positions for junior talent that you will so desperately need to be senior when your current crop retires. You might as well burn the office furniture for heat this winter and save on hvac.


In manufacturing, maybe you haven't done ISO9000 but you've heard of it and have all sorts of regulated documentation as a baseline - a "bias for writing process details down" that is absent in a bunch of other industries, with software at the top of the list. "Documentation // it is written by strangers // throw it in the trash" is a software haiku that I keep running into (not as a policy or anything, just a recurring meme about how bad/uninformative/absent software docs generally are.)




Consider applying for YC's W25 batch! Applications are open till Nov 12.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: