Vicki Boykis walks through how far local LLMs have come: quantized open-weight models now run usefully on a laptop, the tooling (llama.cpp, Ollama, MLX) has matured, and for many tasks you no longer need a frontier API.
Leaked financials show OpenAI burned roughly $3.7B in the first quarter of 2026, with compute and talent costs far outrunning revenue. The numbers sharpen the question of how long the current AI spend can outpace income.
GrapheneOS has finished porting to Android 17, with official releases coming soon — keeping the hardened, de-Googled Pixel OS current within days of the upstream AOSP drop.
In an AP interview, Nvidia's Jensen Huang argued that AI's disruption will require new social norms rather than just regulation, and pledged the technology will create manufacturing jobs — a claim set to be tested at Nvidia's planned Texas buildout.