Most sad thing about Local LLMs:
People are never satisfied.
6 months ago, when running Llama or GPT-OSS on giant servers wasn't even close to frontier models, everyone said they'd gladly use local LLMs even if they only hit last-gen performance.
Now we got 27b models running locally on MacBooks matching Sonnet-4.6, and people still find a way to complain.
The real smart ones just buy the hardware.