For anyone else following what's going on in AI and where we're at, David summarizes it well. 👇

Summary: David Ondrej analyzes the unprecedented government restrictions placed on recent AI model releases. The discussion examines the implications of centralized access to frontier technology, the rise of a permanent digital underclass,

🌟and practical steps individuals can take to host their own models and contribute to decentralized, open-source AI initiatives🌟.

And route them through meshtastic

And while you're at it, build up a few Iranian drones

Buying Hardware for an offline LLM (not Iranian drones 🙄) is easy.

AI Overview (google gemini):

"Running Large Language Models (LLMs) locally on an Apple MacBook Pro is highly efficient due to Apple Silicon's Unified Memory Architecture (UMA). This shared memory setup allows models to utilize massive RAM pools as VRAM, delivering strong privacy, offline access, and zero API costs for heavy local AI workloads.

MacBook Pro AI Hardware Requirements RAM/Unified Memory: 16GB is the bare minimum for smaller models (e.g., Gemma, Phi), while 32GB to 64GB+ is highly recommended for running larger parameter models (like Qwen 3.5 9B) or high-context AI coding tasks.Compute: M-series chips (M4 or M5 series) leverage built-in Neural Engines and optimized Metal frameworks to deliver high token generation speeds."

Simple overview:
1. Buy laptop.
2. Run LLM locally.

AI Power Grab Graphic by Natural News is also a good quick illustration of where we're at now.

[img][/img]