Good morning. It's Sunday, July 12th, 2026.

**Apple sues OpenAI over systematic trade secret theft.** Filed Friday in Northern California, the complaint alleges that OpenAI orchestrated a coordinated recruitment campaign targeting Apple's silicon and on-device AI teams. OpenAI's senior hardware recruiting lead — a former Apple employee — allegedly used internal Apple codenames to extract information from current employees, asked candidates to bring physical prototype components like batteries, logic boards, and system-in-package chips to interviews for "show and tell." Apple also alleges that an eight-year Apple systems engineer who defected to OpenAI this year never returned their company laptop and used it to download confidential files. OpenAI has hired over 400 former Apple employees. Apple is seeking damages and an injunction. This is one of the most dramatic ruptures in Silicon Valley's recent history — a company that integrated ChatGPT into iOS in 2025 now accusing the same partner of theft of trade secrets related to its own hardware ambitions.

**The Federal Reserve puts Marc Andreessen on an AI task force.** Fed Chairman Kevin Warsh announced five external advisory panels this week, and the most eye-catching appointment is a16z co-founder Marc Andreessen, who will co-lead the task force on productivity and jobs — specifically assessing how AI is reshaping employment structures and productivity at the macro level. The panel also includes Stanford economist Chad Jones, who has worked with Anthropic, and Asha Sharma, who leads Xbox at Microsoft. This is the first time the Fed has formally integrated AI as a factor in monetary policy deliberations. For a central bank that has historically treated general-purpose technology shifts — the internet, personal computing — as exogenous, putting a venture capitalist who has bet billions on AI at the table is a signal that the Fed is treating AI-driven productivity as something that could materially affect interest rate decisions.

**The humanoid reality check.** Unitree Robotics just cleared its Shanghai STAR Market IPO with CSRC approval, and Tesla's Fremont assembly line is confirmed building toward a thousand Optimus units per week by September. Both are real industrial milestones. But here's the reality: Unitree's Q1 2026 profit fell 52% year-over-year, and neither company can sell robots to external customers this year. Tesla plans to use its own Optimus units internally first. The stock market surged on Thursday — Hong Kong robotics names jumped 6 to 8% — but the earnings gap between production capacity and actual external revenue is the story. Morgan Stanley doubled its China humanoid deployment forecast to 50,000 for 2026, but the constraint isn't hardware anymore. It's software, deployment integration, and return on investment. The robots exist. The question is whether anyone wants to pay for them at real margins.

**Meta pulls the plug on Instagram's Muse AI image generator after just days.** The feature, launched Tuesday, auto-opted in every Instagram user with a public account and let people generate AI images based on any public profile. Hollywood unions objected. Privacy advocates were outraged. Users complained their likenesses were being used without consent. Meta said on Friday it was discontinuing the feature — admitting it "misses the mark" on privacy expectations. This is a pattern: ship AI features fast, face backlash, walk it back. The question for platforms is whether the cost of these retractions — reputational, regulatory, operational — is starting to outweigh the speed advantage of moving fast.

That's all for today. Have a great Sunday.