Hacker News new | past | comments | ask | show | jobs | submit
I have no interest in having my business or my personal AI usage being reliant on any single third party. We currently have no customer facing LLM products, but if we did they would be only options we can run confidentially which no providers can do ATM to our standards. Current plans are just racking GPUs like I did at home.

Most of my peers do use OpenAI and Anthropic for development work, which pains me, but I will never tell someone else how to do their job if the end result meets my quality bar.

I lead core systems engineering building custom enclaves and operating systems and I do 100% of my work with hardware I own on my own property. When the internet goes down, I do not care and keep building.

If you don't mind me asking, how do you drive the models? Right now it seems like everyone is feeling their way around in the dark about how best to steer the work. AWS AI-DLC, SpecKit, BMAD-METHOD, TwitTech Influencer Pile Of MD files...

It all seems hopelessly adhoc at the moment.

My own fork of jcode stripped down to only my needs in herdr. That is it.