Meta Cooks Up a Real Image Model, Beijing Eyes AI Export Controls Too, DoorDash’s AI Report Card

Today we’re covering the biggest AI stories of July 8th, 2026. Meta’s Superintelligence Labs released Muse Image, its first in-house image model, opening at No. 2 on Arena’s leaderboards behind only OpenAI’s GPT Image 2 and rolling out free across Meta AI, Instagram, and WhatsApp, with a Muse Video model teased right behind it — proof that Alexandr Wang’s earlier claims about catching up to the frontier weren’t just talk. China is reportedly considering its own restrictions on letting top domestic models like Qwen, Doubao, and GLM-5.2 be used outside the country, discussing the idea with ByteDance, Alibaba, and Z AI just weeks after Washington’s Fable and Mythos export controls set the precedent — meaning the world’s cheapest, most popular open models could suddenly become harder for Western users to access too. Plus, DoorDash built an internal benchmark to grade its AI code reviewers and found pairing the free Kimi K2.6 with Claude Fable 5 caught two-thirds of bugs at a fraction of the cost of using either model alone — and today’s community workflow comes from Sarah in London, a 61-year-old van lifer who built an AI decision engine that plans her routes, models her battery life, and helps her choose safe stopovers while living off-grid.