In Today’s Issue:

⚖️ Apple sues OpenAI over stolen iPhone secrets

🥊 Anthropic extends Fable 5 through July 19

🧭 Thinking Machines' manifesto for human-centered AI

🔏 McNealy's "zero privacy" line, 27 years on

🎬 Žižek asks whether AI is becoming a subject

📊 GPT-5.6 Sol tops the DesignArena frontend chart

💰 Zhipu's $4 billion Hong Kong raise

✨ And more AI goodness…

⚡ The Signal

The AI race stopped being polite this week: it is now fought in courtrooms, price lists, and capital markets. Apple suing OpenAI over trade secrets is the clearest signal yet that the next platform war, AI hardware, will be fought with lawyers as much as with engineers. The same logic runs through Anthropic paying to keep Fable 5 in subscribers' hands for another week under GPT-5.6 pressure, and through Zhipu raising $4 billion in Hong Kong with Wall Street banks watching from the sidelines. Model quality decides who gets to play; distribution, capital, and now litigation decide who wins. Watch the lawsuits: they show you where the value sits.

All the best,

Kim Isenberg

Mira Murati's Thinking Machines Lab published its mission manifesto on July 10: AI should multiply human judgment, not replace it. Leaning on the economist Friedrich Hayek, the essay argues that the knowledge running the world is tacit and local, so no centralized model can capture it; the lab wants organizations to own and customize their AI, richer interfaces than chat, and many diverse systems competing instead of one lab setting values for everyone. It is philosophy with a business model attached: Thinking Machines sells exactly that customization layer.

👉 tl;dr: Murati's lab planted its flag: amplify humans with custom, locally owned AI, and own the layer where the customizing happens.

🎬 Watch This

– Anthropic (@claudeai on X, July 12, 2026)

The second extension in six days. Read it as a pressure gauge: GPT-5.6 is pulling hard enough that Anthropic keeps paying, week by week, to hold its best model in subscribers' hands.

A dustup with a twist: days after OpenAI's Codex lead Tibo Sottiaux taunted Anthropic's rate-limit resets with "I smell fear," he spent the weekend shipping fixes for GPT-5.6 Sol's own appetite. In Codex and ChatGPT Work, OpenAI reverted Sol's in-product context window from 372k back to 272k tokens after it drained subscriptions faster than intended, rolled back quiet experiments with reasoning-effort "juice values," and says new inference optimizations should hand subscribers roughly 10% more usage. He calls it "No nerfing, only good stuff!"; users watching their usage meters will judge that for themselves.

The partnership that put ChatGPT inside the iPhone is over, and the fight over what replaces the iPhone has begun. On Friday, Apple sued OpenAI for trade secret theft in the Northern District of California, the sharpest escalation yet between two companies that presented themselves as close partners just two years ago. The suit's language is total: "At every level, from members of its technical staff to its chief hardware officer, and in coordination with business partners, OpenAI has been stealing Apple's trade secrets and confidential information." Apple argues OpenAI's young hardware business is "rotten to its core by its illegal reliance on misappropriated trade secrets."

The complaint centers on Tang Tan, OpenAI's chief hardware officer, who spent 24 years at Apple running product design for the iPhone and Apple Watch. Tan left in 2024 to co-found io Products with Apple design icon Jony Ive; OpenAI bought the startup last year for $6.5 billion. According to Apple, Tan directed job candidates still working in Cupertino to bring "actual parts" to interviews for "show and tell" sessions, and OpenAI "actively coached" departing employees on how to avoid Apple's exit security, including the "dreaded walk out." More than 400 former Apple employees now work at OpenAI, the suit says, among them Chang Liu, who allegedly "surreptitiously accessed and downloaded dozens of Apple's confidential hardware-related files" while already building hardware for OpenAI.

The rupture was a long time coming. Since WWDC 2024, ChatGPT has powered parts of Apple Intelligence and Siri; Apple software chief Craig Federighi once called OpenAI the "pioneer and market leader" in AI. But the alliance soured on both sides: Bloomberg reported in May that OpenAI, unhappy with what the deal delivered, had weighed legal options of its own against Apple. Apple says it tried to resolve the dispute out of court months ago and received no response. Last month the executive running Apple's Vision Pro and smart-glasses work, Paul Meade, also left for OpenAI.

What Apple wants makes the stakes plain: a jury trial, destruction of its proprietary materials, and an order forcing OpenAI to redesign its upcoming products so they contain none of Apple's technology. That remedy aims squarely at the device family OpenAI is building on the io acquisition, and it lands at a delicate moment: OpenAI is preparing an IPO in the coming months, and a discovery-heavy trade-secrets trial is the last thing a listing prospectus wants. OpenAI's answer so far: "We have no interest in other companies' trade secrets. We remain focused on building innovative technology that empowers people everywhere."

Why it matters: Apple v. OpenAI turns the post-smartphone race into a legal war between the company that owns today's device and the company most likely to build tomorrow's. The outcome will set the price of raiding an incumbent's hardware team on the eve of an IPO.

Sources:
🔗 https://www.bloomberg.com/news/articles/2026-07-10/apple-sues-openai-for-trade-secret-theft-in-blockbuster-case
🔗 https://techcrunch.com/2026/07/10/apple-sues-openai-over-alleged-trade-secret-theft/
🔗 https://www.cnbc.com/2026/07/10/apple-openai-lawsuit-trade-secrets.html

Every missed onboarding step, delayed offboard, or broken provisioning handoff has a root cause: HR and IT aren't aligned. This guide gives both teams a shared framework for the full employee lifecycle.

Zhipu, the Beijing lab behind the GLM models, became the first large-model lab anywhere to go public, and six months later it is already back for more capital. Spun out of Tsinghua University, it listed in Hong Kong as Knowledge Atlas Technology on January 8 after a $558 million IPO priced at HK$116.20 a share; the stock has since climbed roughly 1,500%. Its open-source flagship GLM 5.2 sits one Elo point behind GPT-5.6 Sol on today's Graph of the Day, which is a big part of why investors treat Zhipu as China's answer to OpenAI.

On Thursday the company moved to cash in on that run. It is placing 19.8 million new shares at HK$1,588 to HK$1,698 each, up to 13% below Wednesday's close, raising about $4 billion in Hong Kong's second-biggest equity deal of the year. A placement is a fast-track sale of new shares to institutional investors, and the bank list is the quiet headline: state-owned CICC is running the deal as sole coordinator, part of a fundraising wave that Bloomberg notes ran largely without Wall Street banks. The same week, chipmakers Iluvatar CoreX (about $850 million) and Biren (about $900 million) pushed China's one-week AI fundraising total to $5.8 billion.

The skeptic's read: management is selling a large block at a double-digit discount into a euphoric rally, and because Zhipu's free float is tiny, the placement barely changes how much stock actually trades, so the volatility stays. The strategic read is bigger: proceeds are earmarked for R&D, expansion, and M&A, and China's AI champions can now fund compute at scale from Hong Kong's capital market alone. US export controls still limit which chips they can buy; they no longer limit the money.

Hampton co-founder Joe Speiser had three roles budgeted: a data engineer, an ops manager, a PM. $440K. He installed Viktor on April 12. Forty-four days later, none are on the calendar, and 18 of his team work with Viktor daily. His VP: we are editors now, not creators.

Get Started for Free