In partnership with
November 9, 2026 · READ ONLINE
Welcome back, AI enthusiasts.
Nvidia makes the chips behind almost every AI tool we use but today the spotlight is on Jensen Huang after he admitted to an affair amid Chinese spy claims. It is a story about trust and power and how much the AI race still depends on a few people at the top. Let’s Dive In!
IN TODAY'S ISSUE
01 NVIDIA CEO admits affair Suspected Chinese Spy
02 24 Chinese Firms Cleared for Nvidia’s Chips
03 Anthropic’s AI now DREAMS like Humans
01 / NVIDIA ESPIONAGE
Source: NetThunder
Evolving AI: Nvidia is investigating its own CEO after Jensen Huang admitted to an affair.
KEY POINTS
Jensen Huang told employees in a company-wide email he was unfaithful to his wife.
Nvidia’s board is probing claims that Huang was a Chinese intelligence target but still backs him as CEO.
Three House committee chairs want Huang at a closed-door briefing on November 18, a day after Nvidia’s earnings.
DETAILS
US intelligence officials believe China recruited Alice Huang to win the trust of Jensen Huang while her company Megaspeed bought billions of dollars in Nvidia chips. Megaspeed is also under investigation in the US and Singapore over claims that it helped move restricted Nvidia chips into China. Nvidia spent days insisting their meetings were ordinary business until Huang wrote to employees on Sunday that his actions had hurt his wife and children. Nvidia shares fell up to 6% in overnight trading as China dismissed the spy claims as groundless.
WHY IT MATTERS
Jensen Huang turned an idea sketched at a Denny’s into the $4.5 trillion company the AI boom runs on and his image as a loyal and steady founder became part of Nvidia’s value. Investors and Washington have trusted his judgment almost as much as they have trusted the chips his company sells. That is why this scandal cuts deeper than a personal failing because it raises doubts about the judgment of the one man much of the AI industry depends on. Even a clean verdict from the board may not restore the image he spent three decades building.
You know the problem. Now build the fix.
You've spent years close to a broken industry or process. You know exactly what would fix it. But you’ve never had the time or resources to make it happen.
With Lovable, your expertise is all you need. Describe the solution to your problem — a website, a dashboard, an internal tool — and get professional-grade software back. You make every call on how it works, because you're the one who knows what it needs to do.
Start today and take $5 off your first month.
Discount of $5 applies to first month of any paid plan. You need to sign up via this webpage in order to claim the discount. The discount applies to new and returning users without an active paid plan. Terms apply.
02 / US CHIP BAN
Source: Reuters
Evolving AI: The U.S. cleared Chinese firms with blacklisted ties to handle Nvidia’s AI chips.
KEY POINTS
Baidu, BYD and Tesla’s Shanghai unit are among roughly two dozen firms cleared for Nvidia’s H200.
Reporters visited two of them, New Matrix and Maginfra, and found nothing but empty offices.
Analysts tie Maginfra to sanctioned server maker Inspur, and two more approved firms link to suppliers of China’s military.
DETAILS
Trump let Nvidia sell H200 chips to China in December 2025 as long as the U.S. took a 25% cut and Commerce vetted every Chinese firm. New Matrix still passed that check despite being just 17 months old with owners linked to Wise Road, a fund blacklisted in 2024. The department says it reviews every application carefully, yet both empty-office firms were cleared to build AI servers for unnamed customers, though no deliveries are confirmed.
WHY IT MATTERS
China is holding back mass H200 imports to grow its own chipmakers, yet firms linked to its military still hold U.S. approval for Nvidia chips. Long term that gives Beijing a rising homegrown chip industry and a quiet licensed path to American hardware, with Washington’s stamp making it look legitimate.
Building apps with AI is getting easier, but poor planning can still leave you fixing things repeatedly. Google's new 12-minute tutorial shows how to use two Gemini agents to plan an app's design, map its user experience, and document decisions before writing code. It also shares the prompts so you can try the same workflow on your next project.
TOGETHER WITH DATADOG
Evolving AI: Companies are pouring money into AI tools, models, and infrastructure. But there's one question many still struggle to answer: Is all that AI spending actually delivering a return?
Without clear visibility, costs can quickly spiral out of control, and inefficiencies often go unnoticed until the next cloud bill arrives. The good news? Kevel cut its AWS costs by up to $100,000 per month, proving how much of a difference better cost visibility and optimization can make.
Datadog's free guide explores how businesses can take control of their AI spending, uncover hidden inefficiencies, and connect costs to actual performance.
Here's what you'll discover:
Find out where your AI budget goes: See which models, providers, or teams are driving costs.
Catch expensive surprises early: Identify spending spikes before they turn into major bills.
Get more value from AI: Connect infrastructure costs to AI performance and make smarter investment decisions.
03 / DREAMY CLAUDE
Source: Anthropic
Evolving AI: Anthropic revealed Claude “dreams” between tasks by rereading old logs to fix bad memories.
KEY POINTS
Claude was tested on the video game Pokémon Red, where newer Fable 5 beat four bosses with 10 memory files and older Sonnet 3.7 beat one with 413.
A wrong note sent Claude off the same one-way drop three times, but the dreaming-fixed note got it to the next town.
On merger questions that needed several documents, Claude scored 87% with its own wiki and 73% with plain search.
DETAILS
Anthropic calls the process dreaming and modeled it on how the human brain replays recent experiences during sleep. Claude writes notes while it works and later checks those notes against the original session transcripts for errors. Opus 5.5 scored about 84% on a database test with memory turned on and about 66% with it turned off. Strict rules backfired on stronger models, and telling Opus 5 to record only its failures wiped out much of its memory gain.
WHY IT MATTERS
Claude learning from its own past means two people using the same model will slowly end up with very different assistants. Yours will carry months of lessons about how you work, and that history does not easily move with you if you switch apps. Memory is turning into the thing that keeps users loyal, much like years of photos keep people on one phone. The real choice ahead is which company you trust to hold what your AI has learned about you.
04 / HUMAN OR MACHINE?
One is real. Trust your eyes?
Two images. One photograph. Click the one you think is real.
A
B
05 / QUICK HITS
🤖 Perplexity introduced Teammates, AI agents that manage ongoing work like monitoring inboxes and researching competitors, initially for Max subscribers on the web.
💰 OpenAI reportedly closed an $11 billion funding tranche at a $1.25 trillion valuation, with additional financing still planned. The company has not publicly confirmed the terms.
⚡ Anthropic reportedly committed $3.4 billion to 1.7 gigawatts of gas generators for its AI data centers, according to the Financial Times. The company has not confirmed the deal.
🔬 OpenAI unveiled BioBasicsBench, a 100-task test for AI agents working with lab data, with GPT-6 Astra scoring 66.7% on its highest-effort evaluation.
06 / TRENDING AI TOOLS
🚀 Sanebox - AI-Powered Inbox Assistant which automatically filters out unimportant emails*.
🌍 Frameboy - Lets you explore AI-generated 3D worlds in real time.
🧠 Turnstone - Runs personal AI agents using the AI subscriptions you already have.
🔎 Exa Monitors - Watches the web and alerts your AI only when something changes.
*PARTNER LINKS
BEFORE YOU GO
Worth your four minutes?
Help us make the next edition better.
See you tomorrow,
The Evolving AI team
