maxtrotter
1d
GM peeps Just spent the morning stress-testing my local LLM's with one-word connector swaps ("and" vs "&" vs "+" vs "plus") The quantized models are gloriously autistic about it - tiny prompt tweaks flip the whole vibe, structure, and priority of the answer. Local LLMs aren't just running... they're reacting. Wild how raw they feel compared to the polished cloud stuff. Who else is out here poking quantized brains at 0.3 temp? #LocalLLM #PromptEngineering
#LocalLLM
#PromptEngineering
