New_California

Brief: looking for a well balanced laptop well running Linux and easily capable of local inference for small LLM-s, image generation, etc. It will be used standalone so quality display and keyboard are important.

The perfect laptop (I do not expect to find one): * Light and quite 14"-15" * Gaming laptop midrange+ (because that's best for AI) * Great Linux support * NVIDIA GeForce w/ ~8GB VRAM * 32 GB RAM * 2 NVMe slots (Linux for ML/dev/work, Windows for games) * Bright screen 300-500 nits

I am not looking for a monster workstation - I have the PC for that. I am looking for a balance of Linux, gaming-quality-GPU, and quality mobility.

The Asus ROG Zephyrus G14 (2024) comes very close - except for my critical condition of running Linux well - I can see very mixed opinions for ROG series.

Any tips, however far from my elusive ideal are much welcome!

I am looking for a Python library that generalizes over popular LLM API-s without sacrificing full control (need to use tools, control temp, seed, top-p etc).

Simon Willison llm seems close but seems to focus much on the CLI and not the library, also doesn't seem to give precise control for non-OpenAI models (might be wrong here).

I'm looking for a daily updated, opinionated catalog of best AI models, segregated per function. Both SaaS and open models. I mean functions like speech-to-text (Whisper etc), conversational (GPT-4 and open alternatives), text-to-speech, image generation, etc.

I prefer it to only list a tiny number of top-performing models per category. I do not want a bag of hundreds of models to research through.

Essentially, I want opinionated state-of-the-art tracker, because I can't keep up.

Web 1.0 - platform owner generated content

Web 2.0 - user generated content

Web 3.0 - machine generated content; increasingly hard to to tell apart

Thoughts?