
My fellow AI explorers
Amazon started as an online bookstore. Now it is buying rare books, slicing off the spines, and scanning them into an AI training pipeline in Las Vegas. That is not a metaphor; that is literally the week we are having.
Meanwhile, a Trump-backed crypto firm is quietly funneling Americans toward Chinese AI models the same administration has flagged as national security risks. Add in a $105 billion Nvidia bet, a San Francisco housing market gone full AI gold rush, and an MCP integration that turns Claude into your voice agent manager, and yeah… this one's a lot.
In today’s edition:
📚 Amazon's rare book destruction machine, and why it says everything about the AI data race
🇨🇳 A Trump family crypto firm is bankrolling access to restricted Chinese AI
🎙️ Claude can now manage your ElevenLabs voice agents directly (30-Second AI Play)
30-Second AI Play
🎙️ Manage Your Voice Agents Without Leaving Claude
ElevenLabs just launched a hosted MCP integration inside Claude, which means you can now configure, monitor, and improve your voice and chat agents using plain English, no dashboard hopping required.
Here's how to set it up:
Open the Claude connectors directory and find ElevenLabs. Click install; no API key needed.
Sign in once with OAuth. ElevenLabs hosts the MCP server, so there's nothing to run locally, and you can revoke access anytime from Claude or your ElevenLabs account.
Ask Claude to surface recent conversation topics from one of your agents so you can review performance in plain language instead of scrolling logs.
Ask Claude to duplicate an agent for a new market, telling it to update the voice and first message for the new audience automatically.
Ask Claude to compare configurations across all your agents and flag inconsistencies, or to estimate an agent's expected LLM usage and cost before you ship a change.
💡 Pro Tip: Use the cost-estimation step before every agent update. It's the fastest way to catch an expensive configuration mistake before it hits your ElevenLabs bill instead of after.
Stop letting busywork get in the way of selling
Researching accounts. Building lists. Writing sequences.
There's a better use of your team's time.
Apollo is the AI revenue engine that handles the busywork, so you can stay focused on selling.
Plus, everything you need is in one place:
230M+ verified contacts
AI-powered outreach
Data enrichment
Inbound lead capture
Meeting scheduler
And more
Stop doing busywork and start building pipeline, faster.
With Apollo — the AI revenue engine powering 4M+ users.
Books
Amazon Is Buying Rare Books, Slicing Off the Spines, and Feeding Them to Its AI
Amazon, the company that got its start selling books online, is now destroying rare books to train its AI models. A 404 Media investigation tracked a shipment after slipping a tracking device inside a rare book. It ended up at an Amazon facility in Las Vegas known internally as VGT3, marked with a symbol of a dinosaur holding a book in its claws.
Here's what we know:
Amazon told 404 Media it "purchases books through commercial channels to improve the products and services customers use." No further detail.
The process reportedly involves cutting the spines off physical books so they can be scanned at scale, destroying the original in the process.
Rare and out-of-print books are especially valuable because they were never digitized and never scraped, meaning frontier labs haven't already trained on them.
Text published before the AI era (pre-2022) carries a bonus: zero risk of contamination from AI-generated slop, which is exactly the kind of data that causes model collapse when LLMs train on their own outputs.
Here's the uncomfortable part: every major lab has already vacuumed up the open internet. Anthropic just paid $1.5 billion to settle a lawsuit over training on pirated books.
So the next frontier for training data isn't the web anymore. It's physical libraries, estate sales, and out-of-print archives, and the company willing to buy and destroy irreplaceable objects to get there is the one that started out selling them to you.
🔮 Prediction: Rare and physical media is about to become one of the most contested resources in AI, right up there with chips and energy. Expect a wave of quiet corporate buying sprees at auction houses, estate sales, and used bookstores over the next 12 months, and expect libraries and archives to start locking down access or striking licensing deals of their own before their collections become the next scraped dataset.
Reply and tell us: is destroying a physical book defensible if it makes AI smarter, or is this just data piracy with extra steps?
Politics
A Trump Family Crypto Firm Just Became the Side Door Into Restricted Chinese AI
World Liberty Financial, the crypto company backed by President Trump, is collaborating with WorldClaw, a Hong Kong-based platform that gives users access to AI models built by Chinese companies the same administration has flagged as national security threats.
The facts:
WorldClaw offers roughly 90 AI models and accepts World Liberty's USD1 stablecoin as payment.
A Reuters review found 43 of those 90 models come from Chinese firms, including Alibaba, Baidu, Z.ai (formerly Zhipu AI), DeepSeek, and Moonshot.
The Pentagon has designated Alibaba and Baidu as military-aligned companies. Z.ai sits on the Commerce Department's Entity List, which restricts access to U.S. technology.
The Trump family owns roughly 38% of World Liberty and earns a cut of interest on USD1's reserves, meaning every dollar routed through WorldClaw indirectly benefits the president's family.
Trump's two eldest sons have publicly promoted WorldClaw, and a World Liberty executive advises the platform directly.
Here's the uncomfortable part: this is the same administration that has spent the past year restricting American companies' access to Chinese AI over IP theft and surveillance concerns. A White House spokesperson said there are "no conflicts of interest."
Experts told Reuters that routing user data through Chinese models raises real risks, including government monitoring, output censorship, and malicious code slipping into agent workflows. WorldClaw acknowledges user inputs may be shared with the model providers.
🔮 Prediction: This won't stay a crypto story. Expect congressional Democrats and some China hawks in Trump's own coalition to use WorldClaw as Exhibit A in the fight over AI-related conflicts of interest, and expect more "neutral marketplace" AI routers to get caught in the same bind as U.S. restrictions on Chinese models keep tightening while global demand for cheap, capable Chinese models keeps rising.
Reply and tell us: should AI model marketplaces be required to disclose exactly which country built the model you're talking to, the way food labels disclose country of origin?
Advertise to 190k engineers and CTOs choosing what tools their companies build with.
Other Relevant AI News!
🏡 San Francisco's housing market is having a full-blown AI gold rush, with the median home now selling for $1.7 million and roughly one in three Bay Area sales happening all-cash as newly minted AI millionaires bid up everything in sight.
🪦 Automation startup Relay is shutting down for good next month, and its founder is heading back to Google's Chrome team as VP of Product, betting that the browser is "a perfect place to collaborate with agents."
⚡ Nvidia is backing up to $105 billion in financing for OpenAI's massive new Ohio data center campus, reviving fears of circular AI financing even as Jensen Huang insists "OpenAI will pay the lease."
Golden Nuggets
📚 Amazon is destroying rare books to feed its AI, and it's a sign that physical, undigitized media is the new frontier in the AI data war.
🇨🇳 A Trump family crypto firm is bankrolling access to restricted Chinese AI models, exposing a real contradiction inside the administration's own China policy.
⚡ Nvidia's $105B Ohio bet and San Francisco's AI-fueled housing frenzy show the same story from two angles: an enormous amount of money is chasing this boom, and the ripple effects are showing up everywhere from data centers to open houses.
Would love to hear your thoughts! Send me your thoughts by replying to this email (yes, I read them all :)
Until our next AI rendezvous,
Anthony | Founder of Uncover AI

