3 videos · 1–3
On YouTube: No-Jargon AI (@NoJargonDavid)
EmbeddingGemma 2 is Google's new open embedding model: it maps text, photos, audio and video frames into one map of meaning, so you can search everything by meaning, offline, on your own device.
Mac unified memory vs RTX 5090 VRAM, DGX Spark, and local LLM sizing in one buying guide. Is Apple's unified memory actually VRAM? Capacity and bandwidth side by side, with US prices checked on official pages.
Claude Code burns about 28,000 tokens before it answers your first question. I tested Claude Code, Codex, Oh My Pi and the original Pi on the same machine with the same prompt, and broke down where the system prompt overhead actually comes from.