China's monster model, Apple enters China, and the agent economy arrives
Kimi K3 threatens the frontier, Apple Intelligence lands in China, and DoorDash opens a CLI for AI agents.
Subscribe free →China's Kimi K3 Is Coming for the Frontier With 2-3 Trillion Parameters
Moonshot's upcoming Kimi K3 is reportedly the largest open AI model to come out of China, with a parameter count somewhere between 2 and 3 trillion. The Financial Times reports it's expected to close the gap with Anthropic's Opus 4.8 — which would make this a genuine frontier challenger, not just a benchmark-gaming exercise. Open weights at this scale from a Chinese lab would be a major moment for the global AI landscape.
Why it matters: A 2-3 trillion parameter open model from China, competitive with Anthropic's best, would fundamentally shift who has access to frontier-class AI.
Apple Intelligence Finally Gets Into China — With Alibaba and Baidu on Board
Apple has received regulatory approval to launch Apple Intelligence in China, powered by local AI partners Alibaba and Baidu. The deal had been rumored for over a year, and it clears the path for Apple to compete on AI features in one of its most important hardware markets. Without this, Apple risked selling increasingly feature-incomplete iPhones to hundreds of millions of Chinese users.
Why it matters: It confirms the playbook for Western AI companies in China: localize or sit out, and Apple chose localize.
Google AI Mode Can Now Actually Do Things — Not Just Answer Questions
Google is expanding AI Mode in Search to connect with third-party apps, letting users complete tasks — not just get answers — directly from the search interface. The update pushes AI Mode into agentic territory, where Google isn't just a starting point for your workflow but the place where the work happens. It's Google's clearest shot yet at making Search an action layer, not just an information layer.
Why it matters: If Google successfully turns Search into an agent platform, it threatens every productivity SaaS that currently benefits from Google sending it traffic.
DoorDash Now Has a CLI Because AI Agents Don't Use Apps
DoorDash launched a limited beta of `dd-cli`, a command-line tool that lets developers and AI agents search stores, build carts, and place orders straight from the terminal. This isn't a developer curiosity — it's DoorDash explicitly building for a world where AI agents are the customers. The company is betting that agent-native interfaces are coming fast enough to justify infrastructure investment now.
Why it matters: When a $20B delivery company builds a CLI for AI agents, it signals that agentic commerce is moving from research demo to product roadmap across the industry.
Microsoft Is Training Its Sales Team to Trash OpenAI and Anthropic
Microsoft is reportedly coaching its salespeople to position its in-house AI models as more efficient and cost-effective alternatives to OpenAI and Anthropic — the same companies Microsoft has invested billions in and distributes through Azure. It's a striking move that reveals how seriously Microsoft is taking its own model development as a margin play, even at the cost of awkward partner dynamics.
Why it matters: Microsoft actively undercutting OpenAI in sales conversations is the clearest sign yet that their partnership is evolving into a quiet rivalry.
Neko Health Raises $700M to Bring AI Body Scans to the US
Neko Health, the preventive health startup co-founded by Spotify's Daniel Ek, closed a $700 million Series C led by Lightspeed to expand its AI-powered full-body scanning clinics into the United States, starting in New York. The service combines medical imaging, blood tests, proprietary sensors, and clinician review into a single screening visit. It's a massive bet that preventive medicine delivered through AI analysis is a scalable, VC-sized business.
Why it matters: $700M into a single preventive health clinic chain signals that investors see AI diagnostics as the next consumer health category — not just a niche offering for the worried wealthy.
Quick Hits
- →Former DeepMind researcher Andrew Dai raised at a **$300M pre-seed valuation** with no product yet, betting on visual AI as the next major frontier. TechCrunch
- →OpenAI is now selling a **$230 light-up keyboard** designed for its Codex coding agent — while simultaneously in a legal battle with Apple over hardware trade theft allegations. TechCrunch
- →Meta now **alerts parents** when their teenager discusses suicide or self-harm with its AI chatbot. TechCrunch
- →Google is **renaming NotebookLM to Gemini Notebook** and integrating it directly into AI Mode in Search. TechCrunch
- →Energy IPOs are surging at the fastest pace this century as investors scramble for ways to bet on **AI power demand**. Ars Technica
AI TLDR