@0xSero For a second I thought you spent 10k token credits on this task and assumed a small basement fab to autonomously create the computer from scratch 😂
Google engineer explained how to fine-tune a tiny LLM from 46% to 90% accuracy on your phone in 21 minutes - better than $1500 on-device AI bootcamps.
pick Gemma 270M -> generate synthetic task data -> fine-tune with LoRA -> quantize to int4 -> deploy to Pixel and hit 2000 tokens per second.
That loop is how a 270M model beats a 70B one on your task, running fully offline in your pocket.
Gemma 270M + synthetic data + LoRA + int4 quantization + on-device runtime - that's the stack.
Watch and save it, then fine-tune your own tiny agent tonight.
I think the frontier models will super optimize smaller models for highly specific tasks that load into the general purpose robot for this task, being switched out for other tasks.
You wouldn’t send Einstein grocery shopping but explaining once to a simpler mind what has to be done and then focus on the important stuff
@elder_plinius Could open weight models on torrents scattered all over the world realistically be taken down ? Huge efforts I would assume, right ?
@grok what are the chances of making a once released open source model inaccessible for consumers ?