Shredded Books, Surveillance, GPUs
· The Fluency Briefing
The Fluency Briefing
Your Guide to What's Happening in AI and Why It Matters to You
Wednesday, August 19, 2026
An AirTag hidden inside a rare book just tracked it to an Amazon facility where staff slice bindings and feed pages to AI models. Meanwhile, Nvidia wants Wall Street to treat GPU chips like Treasury bonds, a surveillance company built the police tool it promised it never would, and Cerebras says its new chip runs inference 30x faster than GPUs. Three of these stories are about infrastructure nobody was supposed to see - the fourth is a press release.
Today in AI:
- Your Rare Books Are Being Shredded for AI Training - A 404 Media investigation used a hidden AirTag to trace bulk-purchased rare books to an Amazon facility where staff destroy bindings, scan pages, and feed text to AI models. Pre-2022 books are especially prized because they're guaranteed free of AI-generated content. 9to5mac.com
- Flock Safety Built the Surveillance Tool It Said Was Impossible - WIRED reconstructed code from Flock Safety's new AI policing tool and found it can identify drivers, track movement patterns, and cross-reference 911 logs, arrest records, and Social Security numbers. The company previously claimed its cameras "cannot recognize, identify, or track individuals." Aclu
- Nvidia Wants You to Invest in GPUs Like Real Estate - Nvidia partnered with Apollo, BlackRock, Goldman Sachs, and others to assemble $500 billion in financing that treats compute as an investable asset class. CEO Jensen Huang compared chips to "revenue-generating, long-lived, fungible" assets. Financial historians may want to sit down. theverge.com
- Cerebras Launches CS-4, Claims 30x Faster Inference Than GPUs - Cerebras unveiled its fourth-generation wafer-scale system, built from three WSE-3 Turbo processors. The company says it delivers up to 30x faster inference across small and large models, positioning it as an alternative to GPU clusters for real-time AI applications. Cerebras
- China Lets ByteDance and Tencent Import H200 Chips - ByteDance and Tencent each reportedly received 10,000 Nvidia H200 processors, with approval for up to 100,000 each. China told both companies to keep most chips in Hong Kong rather than the mainland, trying to balance frontier AI competitiveness with protecting its domestic chip industry. Finance Biggo
- AI Hiring Tools Face a Wave of Discrimination Lawsuits - A product manager who applied to thousands of jobs without a single interview is suing Eightfold AI in a class action, arguing automated screening functions as an undisclosed consumer report. The case is one of several challenging AI hiring tools on civil rights and transparency grounds. Outsolve
- Hollow-Core Fiber Startup Raises $22M to Rewire Data Centers - Relativity Networks secured $22 million to deploy fiber that transmits data 30% faster than conventional glass by sending light through a vacuum. The company also landed a $40 million follow-on order from an unnamed hyperscaler betting that microseconds matter at campus scale. techcrunch.com
- AI Drug Discovery Has a Data Problem, Not a Model Problem - Startup Vivodyne argues AI won't cure cancer until models train on living human tissue instead of animal studies and protein databases. Even Anthropic's CEO admitted the "AI cures cancer" claim has become more cliche than credible. Cancerresearch
Today's Takeaway:
Nvidia is assembling $500 billion in Wall Street financing to turn GPUs into an asset class you can securitize, trade, and borrow against - the same week Cerebras launched a chip explicitly designed to make GPU clusters obsolete for inference workloads. That timing isn't accidental; it's a tell. Nvidia's financial play only works if GPUs hold their value the way commercial real estate does, depreciating slowly and generating steady revenue.
But Cerebras is arguing that inference - the part of AI that actually serves customers - doesn't need GPUs at all, and Relativity Networks is betting that the physical layout of data centers will look fundamentally different within a few years. If either bet lands, the "long-lived, fungible" asset Jensen Huang pitched to The Verge starts looking more like a depreciating car than a building.
The uncomfortable question for anyone considering compute-backed securities: what's the residual value of a chip when the architecture around it changes faster than a mortgage term?
🔧 Tool Spotlight
- Gamma - Type a topic, get a polished presentation or website in 30 seconds.
What it does: Gamma generates slide decks, documents, and webpages from a single prompt. Pick a theme, customize layouts, and present, share, or export to PDF/PPT.
Who it's for: Anyone who dreads making slide decks - sales reps, teachers, founders pitching ideas, students.
Try this first: Go to gamma.app, click "Create," type a one-line topic like "quarterly review for my landscaping business," and watch a full deck assemble itself.
Free or paid: Free tier with 400 AI credits. Paid plans start at $10/mo for unlimited generation and analytics.
The Bottom Line
The Pattern: The AI industry's physical layer - chips, fiber, books, cameras - is where the real power consolidation is happening right now, and it's moving faster than the policy or financial frameworks meant to govern it. We've tracked data grabs and infrastructure buildouts before; what's new is that the financial engineering has arrived to lock today's hardware choices into decade-long investment vehicles before alternatives even ship.
Our Call: Nvidia's compute-as-asset-class pitch will face its first serious public skepticism - a downgrade, a critical analyst note, or a major investor pulling back - before October 15, 2026. More likely than not. The Cerebras launch gives skeptics a concrete alternative to point to, and Wall Street loves poking holes in novel asset classes. We'll grade this one in a Friday digest.
Your Move: Read The Verge's breakdown of the Nvidia financing structure - takes five minutes. Then ask whoever manages your company's AI infrastructure budget: are any of our compute contracts structured as leases or financing arrangements rather than subscriptions? If nobody knows the answer, that's your Thursday morning agenda item.
What We're Working On
✨ Founding Cohort Special - 60% Off! - Use code MAF20 to join for just $20/month (regularly $50). Get weekly group sessions & workshops, self-paced courses for all levels, access to tools & templates, challenges with peer feedback, and 24/7 support community. → Join Now
✨ Free 30-Minute AI Consultation - Discover how My AI Fluency can help your business unlock the potential of AI. We'll discuss your goals, explore practical AI opportunities for your industry, and outline clear next steps. → Schedule Free Call
✨ How AI-Fluent Are You? - Test your AI fluency with our interactive quiz. See how you stack up and discover what to learn next. → Take the Quiz
💬 Community | 📞 Book a Consultation | 🌐 Website
Fluently yours, The My AI Fluency Team