Four mini PCs run DeepSeek-V4 locally; Germany probes Meta smart glasses

A setup featuring four chained mini PCs with 512GB of RAM was used to run the DeepSeek-V4 Flash model locally, achieving a speed of 14.61 tokens per second. While not fast, this configuration offers an on-premise AI solution without cloud costs or data privacy concerns, though it comes with a $16,000 price tag and potential for early-adopter bugs. In parallel, Germany has seen a criminal complaint filed against Meta and retailers for selling smart glasses that allegedly violate data protection laws by secretly recording individuals, potentially leading to legal consequences for manufacturers and sellers. AI

IMPACT DIY LLM setups offer data privacy for niche users, while legal actions against smart glasses highlight growing privacy concerns in wearables.

RANK_REASON The cluster discusses a DIY hardware setup for running an LLM locally and a legal challenge against smart glasses, neither of which are frontier releases or significant industry shifts.

Read on dev.to — LLM tag →

• Apollo-Optik
• DeepSeek-V4 Flash
• Fielmann
• GEEKOM
• Germany
• HateAid
• MediaMarkt
• Meta
• Mister Spex
• Ray-Ban Meta Wayfarer Gen 2

• infra
• policy

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →