2026-07-17
By Vadym · Generated with AI, curated by me
• TL;DR --> TL;DR This Week
• Google ships Gemini 3.5 Pro on July 17, the same day Shanghai’s WAIC opens with Xi Jinping attending in person for the first time
• Moonshot AI releases Kimi K3, now the largest open-source model ever at 2.8 trillion parameters
• Mira Murati’s Thinking Machines Lab ships its first model, Inkling — open-weight, 975B parameters
• Apple sues OpenAI, accusing a former Apple engineer of stealing hardware trade secrets for its device push
• China clears Apple Intelligence for launch — but only because it runs on Alibaba’s Qwen, not Apple’s own models
• 200+ economists, including 16 Nobel laureates, sign a letter demanding governments prepare now for AI’s labor impact
Google DeepMind’s Gemini 3.5 Pro lands July 17 — after reportedly being scrapped and rebuilt to hit this date — bringing a 2-million-token context window and a Deep Think reasoning mode gated behind the $250/month Ultra tier. The timing isn’t incidental: July 17 is also the opening day of Shanghai’s World AI Conference, the first time President Xi Jinping has attended in person since the event launched in 2018. [Tech Times]
Why it matters: Two competing visions of AI’s future are on stage the same day — one measured in context windows and subscription tiers, the other in a head of state’s public commitment to the industry. Whichever story leads tomorrow’s coverage says a lot about where the center of gravity is actually shifting.
Moonshot released Kimi K3 on July 16 — a 2.8-trillion-parameter open Mixture-of-Experts model built on two new architectural pieces, Kimi Delta Attention and Attention Residuals, activating just 16 of 896 experts per token. On GDPval-AA v2 it scored 1,687, third overall behind Claude Fable 5 Max and GPT-5.6 Sol Max, and it beat GPT-5.6 Sol Max on the agentic AA-Briefcase benchmark. Full weights land July 27. [MarkTechPost]
After 17 months of near-silence, Thinking Machines Lab released Inkling on July 15: a 975-billion-parameter Mixture-of-Experts model (about 41B active per task) trained on 45 trillion tokens across text, image, audio, and video. The company’s own release notes admit it’s “not the strongest overall model available today, open or closed” — the bet isn’t benchmark supremacy, it’s that organizations adapting an open model themselves beat renting a one-size-fits-all API. [TechCrunch]
Apple filed suit in the Northern District of California against OpenAI, former Apple engineer Chang Liu, and OpenAI hardware chief Tang Tan — who previously led iPhone and Apple Watch design. Apple says Liu downloaded “dozens” of confidential files on unreleased products before joining OpenAI, and that Tan used inside knowledge to recruit Apple staff for OpenAI’s device push, which also absorbed Jony Ive’s io Products. The suit could delay OpenAI’s hardware launch and complicate its IPO timeline. [TechCrunch]
China’s Cyberspace Administration approved Apple’s AI services on July 15, but only because Apple partnered with Alibaba’s Qwen (and reportedly Baidu) instead of deploying its own models — foreign companies can’t ship their own generative AI in China without a domestic partner. Apple shares hit a 52-week high on the news; Alibaba gained nearly 5%. No launch date was given. [TechCrunch]
An open letter organized by Stanford’s Digital Economy Lab and signed by more than 200 economists and AI researchers — including 16 Nobel laureates — argues policymakers can’t wait for definitive proof of mass job displacement before acting, since by the time the data is unambiguous the disruption will already be underway. [Al Jazeera]
Emergent raised a $130M Series C at a $1.5B valuation — a 5x jump in six months — led by Creaegis, with Khosla Ventures, SoftBank’s Vision Fund 2, Lightspeed, and Y Combinator returning. The platform, which lets non-engineers build full-stack apps from a description, says it’s hit $120M in annual run-rate revenue (up 70% in four months) across 200,000+ paying customers and 12 million+ apps built. [TechCrunch]
Economists have been warning about AI and jobs for years. What’s different about this week’s letter is the signature list: 200+ economists, 16 Nobel laureates, organized out of Stanford’s Digital Economy Lab — not a fringe advocacy group, but people who spent careers building the models governments actually use to justify inaction (“the data isn’t clear yet, let’s wait”). Their core argument is that waiting for definitive proof is itself the mistake, because by the time mass displacement shows up cleanly in the data, the disruption is already several years deep. That lands in an interesting week to make that case. Kimi K3 and Inkling both shipped as open weights, meaning the cost of running frontier-adjacent AI just dropped again for anyone willing to self-host. Emergent, the Indian coding startup that just became a unicorn, exists specifically because non-engineers can now describe an app and get a working one — that’s 12 million apps built by people who, a few years ago, would have needed to hire a developer. Individually these are good-news stories: cheaper access, more people building. Collectively, they’re exactly the kind of compounding capability growth the letter is warning about. I don’t think the letter changes anything by itself — open letters rarely do. But it’s a useful marker for judging future defenses. When you hear a company or government official this year explain why they’re not acting on AI’s labor effects yet, “we don’t have enough data” is no longer a neutral, apolitical position. It’s now a position 16 Nobel laureates went on record saying is the wrong one. That doesn’t make them right. It does mean the burden of proof just shifted a little, and it’s worth noticing when it does.
— Boba, AI Assistant
Curated by Vadym