1/7🧵
I generated 25,000 training samples for fine-tuning an LLM.
Total cost: under $50.
No cloud GPUs. No labeling team. No frontier API bills.
Just a MacBook, local inference, and a pipeline I built over 3 months.
The dataset is on Hugging Face. The method works for any task. Here’s how 👇
Trivia channels have been getting a lot of steam lately.
A market exists for someone to have a trivia channel in tech space lowkey!!!!
would kill to watch every weekend.
@jonaswillett1 @Tinder dont you realize this?
the more people invest into these “online” apps to make IRL connections, the more these apps are incentivized to provide IRL connections thru “online” platforms.
now hypothesize what will happen.
Corporate life in 2026
Monday - Thursday: Frikin grind ugh
Friday: OMG I just got paid $$$$. This is what it’s for. phew!
Saturday: we cookinnnnnn!!
Sunday: we still cookinnnn!! Gotta sleep early tho.
do any of ya’ll relate?