Nvidia Board Steps In on Jensen Huang
Plus, 💃 How to fix ChatGPT when its answers start getting worse, Anthropic introduces AI dreaming to help Claude remember better and fix past mistakes, and more!


Hola Decoder😎
If someone forwarded this to you and you want to Decode the power of AI and be limitless, then subscribe now and Join Decode alongside 30k+ code-breakers untangling AI.

🚨 Nvidia launches independent investigation into alleged foreign intelligence effort targeting CEO Jensen Huang
Nvidia's board has ordered an independent investigation into reports that CEO Jensen Huang was targeted by a suspected Chinese intelligence operation, following his admission of an extramarital relationship with a woman allegedly linked to China's intelligence network.

The Shift:
- U.S. officials reportedly raised concerns about Huang's relationship with Alice Huang, an Nvidia customer executive whom an intelligence memo allegedly linked to China's top spy agency.
- Jensen Huang acknowledged the extramarital relationship in a letter to employees, apologizing for his actions after previously describing their meetings as normal business interactions.
- Nvidia's independent directors and external lawyers will investigate potential security risks, although the board says it currently has no reason to believe company secrets were compromised.
- Huang remains Nvidia's CEO as the investigation proceeds, while U.S. lawmakers have requested a private briefing by November 11. China has denied the espionage allegations.
The investigation puts Nvidia's leadership under scrutiny at a time when its AI technology is strategically important to both the U.S. and China. While Huang has acknowledged the personal relationship, there is currently no established evidence that confidential information was shared.

Together with AirOps
Could Your Reporting Get the Wrong Channel Cut?

AI search could influence a buyer long before a sale shows up in your reports. If that influence goes uncredited, you could cut spending on a channel that’s helping customers choose you.
On November 20 at 2 PM ET, get inside the measurement work with Google Meridian creator Hitesh Wadhwani, Paramark CEO Pranav Piyush, and AirOps’ COO Matt Hammel.
They’ll break down the blind spots that could be steering your spend:
- Understand where AI search attribution falls short before treating low tracked results as a reason to cut.
- Learn the three requirements for credible measurement so you can spot what your provider’s reporting is missing.
- Explore how AI search fits into existing marketing mix models alongside the channels competing for your budget.
Take the gaps in your reporting back to your team before they become cuts in your budget.
Can't attend the session live? Register anyway, and you’ll get the recordings in 24 hours.

✨How to fix ChatGPT when its answers start getting worse
Researchers at Microsoft and Salesforce ran 200,000+ simulated conversations across 15 top AI models. When the details of a task arrived bit by bit instead of all at once, performance dropped 39% on average. Once a model takes a wrong turn, it doesn’t recover.

1. Know the warning signs - It repeats a mistake you already corrected, brings back an idea you dropped, or fills gaps with guesses you never confirmed. That’s your cue to start fresh, however short the chat is.
2. Extract a clean handoff - “Write one self-contained prompt that captures where we landed: the goal, every requirement, the decisions we made, and what we ruled out. Use only the latest version of each detail. Leave out abandoned ideas and earlier mistakes.”
3. Check it, then start a new chat - Read the handoff once and delete anything wrong. Then paste it into a brand new chat, where the model sees everything up front with no wrong turns to anchor on.
4. Stop the problem before it starts - Open any big task with:“Before you start, ask me every question you need answered to do this well. Don’t attempt anything until I’ve replied.” The study found models guess early and commit to those guesses. This takes the guessing away.
5. Make it permanent for repeat work - If you do this task every week, paste the cleaned handoff into a Project’s instructions. Every chat in that Project starts with the full picture.
Same model, same question. The difference is how the information arrives.

🧠 Anthropic introduces AI dreaming to help Claude remember better and fix past mistakes
Anthropic has revealed a new AI memory approach called “dreaming,” which allows Claude agents to review, correct, and organize their memories during downtime, helping them retain useful information across tasks without starting from scratch.

The Shift:
- Claude agents can save important information from previous tasks as memory files, allowing them to remember past work and continue projects across different sessions.
- During downtime, Claude reviews its saved memories against previous work logs, identifying mistakes, correcting outdated information, and reorganizing what it needs to remember.
- Anthropic found that newer AI models are better at deciding what information is worth saving, with explaining why memory matters proving more effective than giving rigid instructions.
- In Pokémon tests, Fable 5 earned four badges using just 10 memory files, compared with one badge using 413 files for Sonnet 3.7, while Opus 5.5 completed the game.
Anthropic's research suggests that better AI memory isn't about saving everything, but remembering the right information. By reviewing past work and correcting its own notes, Claude can build on previous experiences without carrying unnecessary details into every task.

Together with Playbook Pro
Unlock Hidden Cash to Add Months of Runway

Growth isn’t about chasing new funding; it’s about making every dollar move faster.
Leading DTC brands have gained months of extra runway by simply reshaping how cash enters and exits their business.
Financial Alchemy is your practical guide packed with real operator insights and ready-to-use templates. Inside, you’ll learn how to:
💰 Flip your cash cycle so revenue arrives well before expenses, real examples show brands securing payment weeks ahead of supplier deadlines.
📊 Stress-test growth plans with the Cash Flow Simulator to uncover an average of $80K in trapped working capital each quarter.
🧮 Speed payback with the Unit Economics Analyzer, cutting CAC recovery to as little as five weeks.
🤝 Negotiate supplier terms using the Supplier Toolkit, with tactics that added 12% operating margin in a single contract cycle.
Get your ebook and turn everyday cash flow into your most dependable growth engine.

🏆 Tools you Cannot Miss:
🥁 Drum2Notes by Klangio – Converts drum recordings, uploaded audio, or YouTube music into drum sheet music, MIDI, and MusicXML files automatically.
💬 Toozi – Simplifies tax filing through text-based conversations, allowing users to manage their taxes without navigating complicated forms.
🔄 n8n v2.46.7 – Builds and automates AI-powered workflows with transparent execution steps, allowing teams to inspect, understand, and troubleshoot every action.
🤖 Sim AI – Provides a centralized platform for building, deploying, and managing AI agents with controls for permissions, spending, and data access.
🎤 RapVid – Transforms uploaded photos into AI-generated rap videos featuring solo performances, duets, or rap battles.

🚀 Quick Hits
🚨 A New York Times investigation found a Chinese firm approved to purchase Nvidia’s H200 AI chips operating from an apparently empty office, raising concerns about US export controls and buyer verification.
💰 OpenAI reportedly reached a $1.25 trillion valuation after raising another $11 billion, with funding potentially reaching $20 billion, as rival Anthropic moves toward an IPO following its regulatory filing.
💾 TSMC is reportedly considering developing next-generation 3D DRAM at Nvidia’s request, potentially competing with Samsung, Micron and SK Hynix as soaring AI memory demand drives supply shortages and higher prices.
☁️ Microsoft is reportedly winding down its AI cloud capacity deal with Oracle, declining to renew a 2023 agreement and reducing rented server capacity as it reassesses external AI infrastructure partnerships.
📱 Meta launched Muse Spaces, letting users create custom AI-powered mini apps through simple prompts, with connected apps providing automatic updates, interactive dashboards and personalized experiences alongside expanded memory capabilities.

👨💻 What’s trending on socials

🎬 A designer showcased a cinema booking app concept that previews the exact screen view from your selected seat, helping moviegoers compare viewing angles and distances before purchasing tickets.
🇩🇰 Denmark has proposed a copyright law granting citizens rights over their faces, voices, and bodies, aiming to protect individuals against unauthorized AI-generated deepfakes and digital impersonation.
🎥 Microsoft has reportedly filed a patent for AI that rewrites speech during livestreams in real time, potentially removing profanity, personalizing language for viewers and synchronizing lip movements with altered dialogue.
💼 Anthropic CEO Dario Amodei warned that AI can already perform entry-level work in law, finance, and consulting, raising concerns about shrinking career opportunities for graduates entering the workforce.
📉 Reddit discussions show growing negativity toward AI, with negative comments rising from 37% to 45% across four technology communities, while ChatGPT- and Claude-focused communities showed improving sentiment despite temporary declines.

🤳AI Nugget of the Day


Thanks for Decoding with us🥳
Your feedback is the key to our code! Help us elevate your Decode experience by hitting reply and sharing your input on our content and style.
Keep deciphering the AI enigma, and we'll be back with more coded mysteries unraveled just for you!
