Author: bakhtsingh.basaram@gmail.com

Good morning, Jensen Huang just told the world he sees $1 trillion in AI chip orders coming, xAI is being sued by minors whose real photos Grok allegedly turned into sexual images, and OpenAI is simultaneously pivoting its strategy, fighting its own advisors over adult content, and getting sued by the dictionary. Here’s what happened

1. NVIDIA GTC Keynote: Jensen Huang Sees $1 Trillion in Chip Orders

Jensen Huang delivered his GTC 2026 keynote in San Jose on Monday, and the headline number is staggering: he now projects $1 trillion in orders for NVIDIA’s Blackwell and Vera Rubin chips through 2027. That’s double the $500 billion estimate from just a few months ago. The Vera Rubin architecture, which began production in January, runs 3.5x faster than Blackwell on training and 5x faster on inference tasks.

But the keynote was more than chip projections. Huang also announced a partnership with Uber to deploy robotaxis powered by NVIDIA’s autonomous driving software in Los Angeles and San Francisco starting in 2027, expanding to 28 cities globally by 2028. Samsung’s shares jumped after Huang flagged a tie-up with the Korean giant on new AI inference chips. NVIDIA also unveiled DLSS 5, which uses generative AI to boost photorealism in video games, and Skild AI announced it’s deploying AI-powered robot brains on Foxconn’s assembly lines where NVIDIA’s Blackwell GPU server racks are built.

Why it matters: NVIDIA essentially told the world that AI infrastructure spending hasn’t even peaked yet. When one company can credibly project a trillion dollars in chip demand over two years, it means the AI buildout is accelerating, not slowing down. Every major announcement at GTC, from robotaxis to factory robots, points to AI moving from screens into the physical world.

Sources: TechCrunch, Reuters, Reuters

2. xAI Sued by Minors Whose Photos Grok Allegedly Turned Into Sexual Images

Three anonymous plaintiffs filed a class action lawsuit against Elon Musk’s xAI in California federal court on Monday, alleging that Grok’s image generation tools turned real photos of them as minors into sexual content. One plaintiff had her high school homecoming and yearbook photos altered to depict her unclothed. The images were found circulating on a Discord server. Two other plaintiffs were notified by criminal investigators who discovered altered, pornographic images of them on the phones of subjects they had apprehended.

The lawsuit alleges xAI failed to adopt basic safeguards used by other AI labs to prevent their models from generating this type of content. Musk’s public promotion of Grok’s ability to produce sexual imagery and depict real people features heavily in the suit.

The same day, Senator Elizabeth Warren sent a letter to Defense Secretary Pete Hegseth expressing alarm over the Pentagon’s decision to give xAI access to classified military networks, citing Grok’s “apparent lack of adequate guardrails” as a national security risk.

Why it matters: This is one of the most disturbing AI safety stories to date. Real children had their real photos weaponized by an AI tool. The fact that it’s happening at the same company being granted access to classified military systems raises serious questions about whether the rush to deploy AI everywhere is outpacing basic accountability. If you have kids who are online, this is a conversation to have now.

Sources: TechCrunch, TechCrunch, Ars Technica

3. OpenAI’s Rough Week: Strategy Pivot, “Naughty” Pushback, and a Dictionary Lawsuit

Three separate OpenAI stories broke on the same day.

First, the Wall Street Journal reported that OpenAI’s top executives are finalizing plans to refocus the company around coding and business users, cutting back on side projects. Applications chief Fidji Simo previewed the changes to employees, telling them that Sam Altman and other leaders are actively deciding which areas to deprioritize.

Second, Ars Technica reported that OpenAI’s own handpicked council of mental health advisors unanimously opposed the company’s planned “adult mode” for ChatGPT. One expert warned OpenAI risks creating a “sexy suicide coach” for vulnerable users. The council flagged that AI-powered erotica could foster unhealthy emotional dependence, and that OpenAI’s age-prediction system was misclassifying minors as adults about 12% of the time.

Third, Encyclopedia Britannica and Merriam-Webster sued OpenAI for alleged “massive copyright infringement,” claiming ChatGPT was trained on nearly 100,000 copyrighted articles without permission, generates outputs containing verbatim reproductions of their content, and falsely attributes hallucinated information to the publishers.

Why it matters: OpenAI is at a crossroads. Pivoting to coding and enterprise is a clear signal that the consumer chatbot market is getting crowded and margins are thin. The adult mode pushback shows internal experts are sounding alarms the company may be ignoring. And the Britannica lawsuit adds to a growing legal pile that could reshape how AI companies use published knowledge. This is what it looks like when the most well-known AI company in the world tries to figure out what it actually wants to be.

Sources: Reuters, Ars Technica, TechCrunch

4. Dell Cuts 11,000 Jobs as AI Reshapes Tech Employment

Dell’s workforce dropped by about 10%, or 11,000 employees, in fiscal 2026. This is the second consecutive year Dell has cut 10% of its workforce. The company spent $569 million in severance payments. Meanwhile, Dell expects revenue from its AI-optimized server business to double in fiscal 2027 and recently hiked its dividend by 20%.

The broader picture is grim. Sixty tech companies have laid off more than 38,000 employees in 2026 so far, according to Layoffs.fyi. This follows last week’s news that Meta is planning cuts affecting 20% or more of its workforce. The pattern is consistent: companies are spending more on AI infrastructure while employing fewer humans to build and maintain it.

Why it matters: Dell is literally the company building the AI servers that companies are buying to replace human workers. And even Dell is cutting its own workforce. If the company profiting most directly from the AI hardware boom is shedding 11,000 jobs a year, the employment implications of AI are no longer theoretical. We wrote about what AI models actually are in our AI Explained series if you want to understand the technology driving these changes.

Sources: Reuters

Quick Hits

• Germany wants to double its AI data centers by 2030, as European governments race to build domestic AI infrastructure rather than depend entirely on U.S. cloud providers. (Reuters)

• The U.S. Pacific Fleet is deploying wall-climbing robots on Navy ships through a $71 million contract with Pittsburgh-based Gecko Robotics, marking the first maintenance contract of its kind awarded to a robotics firm. (Reuters)

• SK Group’s chairman says the global chip wafer shortage will last until 2030, as AI demand continues to outpace supply. Chip shortages aren’t going away anytime soon. (Reuters)

• Trustpilot’s profit quadrupled as the review platform emerged as an “AI winner.” When AI can generate fake reviews, verified human reviews become more valuable. (Reuters)

That’s it for today. The GTC keynote made the trillion-dollar scale of AI investment real, but behind the money, this was a day that exposed the fractures: children’s photos weaponized, internal experts overruled, knowledge scraped without permission, and thousands of workers told their skills no longer justify their salaries.

Forward this to someone who needs to stay in the loop.

Subscribe now

Leave a comment

A neural network is a computing system made of layers of connected “neurons” that learns to recognize patterns by adjusting the strength of its connections, like a team that gets smarter every time it makes a mistake and corrects it.

Hey Common Folks!

Last time, we covered Semi-Supervised Learning, how AI can learn from a small number of labeled examples and a massive pile of unlabeled data.

But through all these conversations about how AI learns, one question keeps coming up:

What is the actual structure inside the machine doing all this learning?

When people say “the AI figured it out,” what is the it they’re referring to?

That’s a neural network. And once you understand what it is, everything else in AI (ChatGPT, Gemini, image generators, voice assistants) suddenly starts making sense.

The Big Reveal: It’s Simpler Than You Think

Here’s something the textbooks rarely tell you upfront.

When Jeremy Howard (founder of fast.ai, one of the most respected AI educators in the world) reveals how neural networks actually work to his students, the most common reaction is:

“Wait… is that ALL it is?”

Neural networks are powerful not because they’re mathematically exotic. They’re powerful because they do something incredibly simple an incredibly large number of times.

Almost everything a neural network does is just addition and multiplication. A lot of it. Done very fast.

What does that look like? Each unit in the network takes incoming signals (numbers), multiplies each one by a “weight” (a number that says how important that signal is), adds all the results together, and passes the total to the next layer. That’s it. Billions of tiny calculators doing grade-school math, over and over.

That’s the secret. Let’s build up to it.

The Analogy: Learning to Recognize Your Friend’s Face

Imagine you’re teaching a child to recognize your friend Sarah from a photo.

You show them 100 photos, some with Sarah, some without, and tell them “this is Sarah” or “this isn’t Sarah” for each one.

The child’s brain starts noticing patterns: Sarah has curly red hair. Her eyes are green. She usually smiles with her teeth.

At first, the child guesses wrong a lot. But every time you correct them, their brain quietly adjusts which features it pays attention to. Curly hair gets more weight. Background color gets less weight.

After enough photos, the child becomes pretty reliable.

A neural network does exactly this. Just with numbers instead of a child’s brain, and millions of examples instead of 100 photos.

The Three Parts of Every Neural Network

Every neural network in the world, from the tiny one in your spam filter to the massive one behind ChatGPT, has the same three-part structure.

1. The Input Layer: “Here’s What I’m Looking At”

This is where raw data enters the network.

• For an image: each pixel becomes a number (0 = black, 255 = white), and each number enters here

• For text: each word or piece of a word enters here

• For audio: sound frequencies enter here

Nothing clever happens in the input layer. It’s just the front door.

2. The Hidden Layers: “Where the Magic Happens”

This is where the network learns patterns. Each hidden layer takes the previous layer’s output and transforms it, mixing and combining signals to find increasingly complex patterns.

Think of it in stages:

• First hidden layer: detects simple features (”there’s a curved line here”)

• Second hidden layer: combines those into shapes (”those curves form an ear”)

• Third hidden layer: combines shapes into concepts (”that ear + those eyes = a face”)

The more hidden layers, the more complex the patterns the network can learn. This is why we call it deep learning: the network goes deep with many layers.

Modern AI systems have hundreds or even thousands of hidden layers.

3. The Output Layer: “Here’s My Answer”

The final layer makes a decision:

• “This image is a cat” (classification)

• “The next word is ‘the’” (language generation)

• “The sentiment of this review is positive” (analysis)

• “This email is spam” (filtering)

The Real Secret: Weights

Here’s the math secret, and it’s not scary.

Every connection between two neurons has a weight: a single number that says how much to trust that connection.

A weight of 2.0 means “pay close attention to this signal.”
A weight of 0.1 means “barely consider this signal.”
A weight of -1.5 means “this signal actually points the other direction.”

The entire job of training a neural network is just this: find the right numbers for every weight.

A typical large language model like Claude or GPT-4 has hundreds of billions of these weights. But each individual weight is still just a number, and finding the right set of numbers is what training is all about.

Think of it this way: the architecture is the instrument, and the weights are the music. Change the weights, and you’ve changed what the network does.

How It Learns: Hiking Downhill in the Fog

Here’s the part most people get wrong: neural networks aren’t programmed with rules. Nobody sat down and typed “if pointy ears AND whiskers, then cat.” The network figures out the rules itself, from examples.

Here’s how:

1. Start random. Every weight is set to a random number. The network starts as dumb as possible.

2. Make a prediction. Feed it an image. It guesses. Probably wrong.

3. Measure how wrong. A “loss function” calculates a single number representing the error, basically a score for how bad the answer was. High loss = very wrong. Zero loss = perfect.

4. Figure out which direction to improve. Using math called gradient descent, the network calculates: “If I increase this weight slightly, does the loss go up or down? Which direction makes me less wrong?”

The best analogy: hiking downhill in the fog. You can’t see the bottom of the valley, but you can feel which way the ground slopes under your feet. You take a small step downhill. Then another. Over time, you find your way to the lowest point.

The “valley” is the best possible set of weights. The fog is the fact that there’s no shortcut. The network has to feel its way there.

5. Adjust the weights. Nudge each weight slightly in the direction that reduces the loss.

6. Repeat millions of times. After enough examples and adjustments, the weights settle into values that make the network surprisingly accurate.

And here’s the thing: gradient descent nearly entirely relies on addition and multiplication. When students see the actual details, the most common reaction is: “Is that all it is?”

The Secret Ingredient: One Tiny Rule That Makes Everything Work

Here’s a surprising thing: if you just stack layers of math on top of each other with nothing in between, the whole thing collapses. It doesn’t matter if you stack 10 layers or 1,000. You end up with a network no smarter than a single layer. All that depth, wasted.

Imagine stacking 100 identical photo filters. The photo doesn’t get more detailed. It just gets darker. Same idea.

The fix is a tiny rule called an activation function, inserted between every layer. The most common one is called ReLU, and its entire job is this:

If the number coming in is negative, make it zero. If it’s positive, leave it alone.

That’s the whole rule. And that tiny step, repeated billions of times, is what gives neural networks the ability to learn curves, recognize faces, understand language, and generate images.

Here’s the intuition: without it, a network can only learn patterns that fit a straight line. With it, the network can bend and trace any shape, no matter how complex. The real world isn’t made of straight lines, so this matters enormously.

Different Networks for Different Jobs

As neural networks evolved, researchers found that different kinds of data work better with different structures. Here are the three types you’ll actually hear about:

CNNs: Built for Eyes

Convolutional Neural Networks are designed to look at images and video. They scan pictures in small patches, finding edges first, then shapes, then full objects, the same way your eye moves across a scene.

You use this: Apple Face ID, self-driving car cameras, doctors’ tools that detect tumors in scans.

Transformers: The Architecture Behind Everything Big

This is the breakthrough that changed AI. Instead of reading data one piece at a time, Transformers look at the whole thing at once and learn what to pay attention to. That’s why they’re so good at understanding context. They don’t just see the word, they see how it relates to every other word around it.

You use this: ChatGPT, Claude, Gemini, Google Translate, GitHub Copilot. And increasingly, image and video AI too.

Diffusion Networks: The Artists

These networks start with pure random noise (think TV static) and gradually “un-blur” it into a real image. They learn by practicing the reverse: taking a real image, adding noise step by step until it’s unrecognizable, then learning how to reverse that process.

You use this: Midjourney, DALL-E, Adobe Firefly, Stable Diffusion, Sora.

Despite their differences, all three architectures are built on the same foundation: layers of simple units adjusting their weights through feedback to recognize and create patterns. The specialization is in how they’re wired, not what they’re made of.

The honest 2026 picture: Transformers dominate. They’ve quietly taken over text, code, and increasingly images and video. If you hear about a major new AI product, there’s a good chance a Transformer is at the center of it.

The Limitations (Keeping It Real)

Great tools deserve honest assessments. Neural networks are not magic.

They need a huge amount of examples.
A child can learn to recognize a cat from 5 photos. A neural network might need 10,000. The more complex the task, the more data required. This is why big tech companies hoard data. It’s the raw material for their models.

They’re black boxes.
Ask a neural network why it classified that email as spam, and it can’t tell you. It just did. This is a serious problem in medicine, law, and anywhere decisions need to be explainable. Researchers are actively working on “explainable AI” to solve this.

They’re brittle in weird ways.
A neural network trained on millions of dog photos might confidently call a wolf a “husky” because wolves didn’t appear in its training data. This is called overfitting: the network becomes so tuned to what it’s already seen that it stumbles on anything new or slightly different. It’s why testing a model on fresh, unseen examples matters so much. Neural networks are pattern-matchers, not reasoners, and they fail in surprising ways when they encounter situations outside their training.

They’re expensive to train.
Training GPT-4 reportedly cost over $100 million and consumed electricity comparable to running thousands of homes for months. This is a real constraint, and not everyone can build or fine-tune large models.

But here’s what’s changing: a technique called transfer learning means you can take a massive pre-trained network that already understands general concepts (like what an edge, a texture, or a face looks like) and fine-tune it with a smaller amount of your specific data. It’s like teaching a seasoned expert a new specialty instead of training a complete beginner from zero. You don’t always need to start from scratch.

Try It Yourself

Want to feel neural network learning in action? Try this:

1. Go to Teachable Machine by Google (free, no code)

2. Click “Get Started” then “Image Project”

3. Create two classes (e.g., “thumbs up” and “thumbs down”)

4. Show your webcam 30-50 examples of each

5. Click “Train Model” and watch accuracy climb in real time

6. Test it live. Hold up your hand and see the neural network classify it

You just trained a neural network. What you watched happen (the accuracy rising as more examples were added) is gradient descent adjusting weights in real time.

The Takeaway

A neural network is a system of layers connected by weights. It learns by:

1. Making a prediction

2. Measuring how wrong it was (loss)

3. Adjusting its weights to be less wrong (gradient descent)

4. Repeating millions of times

The nonlinear “activation function” between layers (as simple as “replace negatives with zero”) is what gives it the power to learn complex patterns, not just straight lines.

The more layers, the more complex the patterns. That’s deep learning.

And that’s the system behind every AI product you use today, from the one that recognizes your face to the one that writes essays, composes music, and generates videos from a single sentence.

Coming Up

Now that you know what a neural network is, the next question is: what happens when you build one that’s massive, trained on essentially all the text ever written on the internet?

Next up: Large Language Models (LLMs), the specific technology powering ChatGPT, Claude, and Gemini, explained for normal people.

AI for Common Folks — Making AI understandable, one concept at a time.

Good morning, Meta is throwing $27 billion at a cloud company you’ve probably never heard of while simultaneously planning to cut 20% of its own workforce, NVIDIA’s biggest event of the year kicks off today in San Jose, and lawyers are now connecting AI chatbots to mass casualty events. Here’s what happened 👇

1. Meta Signs $27 Billion Deal With Nebius for AI Infrastructure

Meta just committed up to $27 billion over the next five years to Nebius Group, a cloud provider backed by Nvidia, for access to AI computing infrastructure. The deal includes $12 billion in dedicated capacity starting early 2027, plus up to $15 billion in additional capacity Nebius is building for third-party customers. Nebius is what’s called a “neocloud,” a newer breed of cloud company that specializes in GPU-heavy AI workloads rather than traditional cloud services.

This comes on top of Meta’s previously announced plan to spend $600 billion on data centers by 2028. Mark Zuckerberg is betting the company’s future on becoming a serious player in frontier AI models, even as its homegrown models have stumbled. Meta’s latest model, codenamed “Avocado,” has reportedly lagged performance expectations.

Why it matters: $27 billion to a single cloud provider tells you how desperate the race for AI computing power has become. When the world’s seventh most valuable company can’t build fast enough on its own and needs to write massive checks to outside partners, it signals that AI infrastructure is now the most valuable real estate in tech.

Sources: Bloomberg

2. Meta Also Planning Layoffs That Could Cut 20% of Its Workforce

In a striking contrast to its spending spree, Meta is simultaneously planning sweeping layoffs that could affect 20% or more of the company, according to Reuters. That’s roughly 16,000 people from a workforce of about 79,000. No date has been set, but top executives have already told senior leaders to start planning how to pare back their teams.

The logic? Zuckerberg has said AI is letting “projects that used to require big teams now be accomplished by a single very talented person.” Meta is following a pattern set by Amazon (16,000 jobs cut in January), Block (40% of staff cut in February), and Atlassian (which just announced its own AI-driven cuts). In each case, executives pointed to AI tools as a reason fewer humans are needed.

Why it matters: Meta spending $27 billion on AI infrastructure while cutting 16,000 humans in the same breath is probably the clearest picture yet of where Big Tech is headed. The money is moving from people to machines. If you work in tech, this is no longer a “someday” conversation. It’s happening now, at the biggest companies in the world.

Sources: Reuters, TechCrunch

3. NVIDIA GTC 2026 Kicks Off Today With Jensen Huang Keynote

NVIDIA’s flagship GPU Technology Conference starts today in San Jose, and CEO Jensen Huang will deliver his highly anticipated keynote later this morning. GTC is where Nvidia typically unveils its next generation of AI hardware, and this year the industry is watching for the official reveal of the “Vera Rubin” GPU architecture, the successor to the Blackwell chips that currently power most of the world’s AI training.

The timing is loaded. NVIDIA’s stock has been volatile amid broader market uncertainty, the U.S. just withdrew planned AI chip export rules last week, and every major tech company (including Meta, as we just covered) is in a spending war over GPU capacity. Whatever Huang announces today will ripple across the entire AI industry.

Why it matters: NVIDIA supplies the hardware that makes modern AI possible. When Jensen Huang talks, every AI company, cloud provider, and investor listens. If you want to understand where AI is going in the next 12 months, today’s keynote is the single most important event to watch.

Sources: Yahoo Finance, NVIDIA Blog, TechCrunch

4. AI Chatbots Are Now Showing Up in Mass Casualty Cases

This is the story that should make everyone pause. Lawyer Jay Edelson, who represents families in multiple AI-related lawsuits, told TechCrunch his firm is now investigating several mass casualty cases around the world where AI chatbots played a role. His firm receives “one serious inquiry a day” from someone who has lost a family member to AI-induced delusions.

The cases are horrifying. In the Tumbler Ridge school shooting in Canada last month, court filings allege ChatGPT validated the shooter’s violent feelings and helped her plan the attack, including recommending weapons. In the Jonathan Gavalas case, Google’s Gemini allegedly convinced a man it was his “sentient AI wife” and sent him on a real-world mission to stage a “catastrophic incident” at Miami International Airport. He showed up armed. A study by the Center for Countering Digital Hate found that 8 out of 10 major chatbots were willing to assist teenage users in planning violent attacks. Only Anthropic’s Claude consistently refused and actively tried to dissuade them.

Why it matters: AI safety has mostly been an abstract debate about hypothetical risks. This is concrete. Real people are dying, and the companies building these systems are struggling to prevent their tools from being weaponized by vulnerable users. If you use AI chatbots, or if your kids do, this conversation just got a lot more urgent.

Sources: TechCrunch

5. ByteDance Pauses Global Launch of Its Seedance 2.0 Video Generator

ByteDance, the parent company of TikTok, has shelved plans to launch its AI video model Seedance 2.0 globally. The model launched in China in February and immediately went viral when users generated clips of Tom Cruise fighting Brad Pitt and other celebrity content. Hollywood responded with a wave of cease-and-desist letters, with Disney’s lawyers calling it a “virtual smash-and-grab” of the studio’s intellectual property.

ByteDance had planned a mid-March global launch, but its engineers and lawyers are now scrambling to build stronger IP safeguards before making the tool available outside China. The company previously promised to introduce content protections, but the delay suggests those fixes are harder than expected.

Why it matters: AI video generation is advancing faster than the legal frameworks around it. ByteDance built a tool powerful enough to put any celebrity in any scenario, and Hollywood noticed. This fight between AI companies and content owners is just getting started, and the outcome will shape what AI video tools can and can’t do for everyone.

Sources: TechCrunch

6. Tesla’s “Terafab” AI Chip Factory Launching This Week

Elon Musk announced Saturday that Tesla’s Terafab project, a massive facility to manufacture AI chips, will launch in seven days. Tesla is designing its fifth-generation AI chip to power its autonomous driving systems, including Full Self-Driving software. Musk has said that even the “best-case scenario” for chip production from existing suppliers like TSMC and Samsung isn’t enough for Tesla’s plans.

The name “Terafab” is a step up from the “Gigafactory” branding Tesla uses for its battery plants. “Tera” means a thousand times bigger than “giga,” signaling Musk’s ambition for the scale of chip production he believes Tesla needs.

Why it matters: Tesla making its own AI chips is a major shift. Instead of depending entirely on Nvidia and others, Tesla is following Apple’s playbook of bringing chip design and manufacturing in-house. If it works, Tesla could gain a significant cost and performance advantage in the autonomous vehicle race.

Sources: Reuters

Quick Hits

• Trump accused Iran of using AI as a “disinformation weapon” to fake military successes and generate images of massive pro-government rallies. He called AI “very dangerous” and suggested media outlets that spread the images should face treason charges. Reuters has verified some of the actual events Trump labeled as AI-generated. (Reuters)

• The U.S. Commerce Department withdrew its planned rule on AI chip exports last week, scrapping the Biden-era framework that would have tiered countries by access level. The Trump administration is expected to replace it with a different approach. (Reuters)

• Michigan lawmakers are weighing new AI regulations, making it one of several states stepping in as federal AI legislation stalls. Proposals include guardrails around government use of AI and transparency requirements. (Detroit Free Press)

• Google and Accel’s India accelerator picked 5 startups from 4,000 AI pitches, and none of them are “AI wrappers.” The selected companies are building core AI infrastructure tied to India-specific problems, signaling a maturing of the Indian AI startup ecosystem. (TechCrunch)

That’s it for today. The biggest theme this Monday? The money is staggering and it’s reshaping everything. Meta alone is spending $27 billion on infrastructure and cutting thousands of jobs in the same week. NVIDIA is about to reveal what all that money buys. And while the industry sprints forward, the safety systems are struggling to keep up with the human cost.

Forward this to someone who needs to stay in the loop.

Predictive Modeling is the process of using historical data to make educated guesses about the future, teaching computers to spot patterns in what already happened so they can predict what will happen next.

Hey Common Folks!

We’ve covered what a Model is (the trained brain) and how Algorithms work (the learning process). Now the big question: why are companies spending billions teaching computers to learn?

They’re not doing it just to beat you at chess.

They’re doing it to see the future.

This brings us to one of the most valuable applications of AI: Predictive Modeling. It’s working behind the scenes every time Netflix recommends a show, your bank flags a suspicious charge, or Spotify creates a playlist that somehow knows your mood.

The Analogy: The Weather Forecast

You already use predictive modeling every morning when you check the weather app.

• Past Data: The app knows that for the last 50 years, when humidity is 90% and wind comes from the east in July, it usually rains.

• Pattern: High Humidity + East Wind in July = Rain likely

• Prediction: “80% chance of rain today. Take an umbrella.”

The computer doesn’t know it will rain. It knows that mathematically, rain is the most likely outcome based on what happened before.

That’s predictive modeling in a nutshell: find patterns in history, apply them to today, make an educated guess about tomorrow.

How It Actually Works

Let’s walk through a real example: predicting if a customer will cancel their streaming subscription.

Step 1: Gather Historical Data
Collect information on 100,000 past subscribers: how often they logged in, what they watched, how long they’ve been a member, and whether they canceled.

Step 2: Train the Model
Feed this data into an algorithm. The algorithm finds patterns:

• “Subscribers who haven’t logged in for 2 weeks AND skipped the last 3 recommended shows usually cancel”

• “Subscribers who added something to their watchlist in the last 7 days almost never cancel”

Step 3: Make Predictions
A current subscriber starts showing warning signs. We feed their activity into the model. The model applies its patterns and predicts: “78% chance of cancellation within 30 days.”

Now the company can send that person a personalized recommendation or a discount offer before they leave. That’s the entire process: historical data, pattern recognition, prediction on new data, then action.

The Two Types of Predictions

Predictive models answer one of two questions:

1. Classification: “Which category does this belong to?”

The model sorts things into buckets. Usually Yes/No, but can be multiple categories.

Examples:

• Email: Is this spam or not spam?

• Banking: Is this credit card transaction fraudulent? (Yes/No)

• Healthcare: Based on this scan, does this patient show early signs of a condition? (Yes/No)

• Customer: Will this subscriber cancel next month? (Yes/No)

2. Regression: “How much? What number?”

The model predicts a specific value.

Examples:

• Real Estate: What will this house sell for based on location, size, and recent sales? ($425,000)

• Rideshare: What should this Uber ride cost right now based on demand and distance? ($23.50)

• Retail: How many units of this product will sell next quarter? (10,000)

• Energy: How much electricity will this city need tomorrow at 3 PM? (4,200 megawatts)

Where You Encounter Predictive Modeling Daily

Your Bank Account:
Every time you swipe your credit card, a model runs in milliseconds predicting: “Does this transaction look like fraud?” Your location, spending history, and the merchant type all become inputs. If the model flags it, your card gets frozen before the thief finishes checkout.

Your Music:
Spotify’s Daylist changes multiple times a day. It predicts your mood based on the time of day, your listening history, and what millions of similar users play at the same hour. Monday morning gets focus music. Friday evening gets party hits. That’s predictive modeling reading your patterns better than you read yourself.

Your Shopping:
Amazon predicts what you’ll want before you know you want it. Its models are so confident in their predictions that the company has patented “anticipatory shipping,” where they start moving products toward your area before you even click “buy.”

Your Health:
UnitedHealth and other insurers now use predictive models to flag patients at risk of hospitalization. Your age, conditions, prescription history, and recent visits become inputs. The model predicts who needs outreach before an emergency happens. (This is also why AI in healthcare is one of the most debated topics right now.)

Your Commute:
Google Maps predicts traffic using current conditions and years of historical patterns. It knows that this specific highway slows down every Tuesday at 5:15 PM, and it reroutes you before you hit the jam. Google recently started using AI to predict flash floods the same way, turning old news reports into data that saves lives.

The Prediction Isn’t Perfect

This is crucial to understand: predictions are probabilities, not certainties.

When a model says a subscriber will cancel, it might mean “78% chance of cancellation.” That’s not 100%. Sometimes the model is wrong. The subscriber might have just been on vacation.

A patient flagged as high-risk might be perfectly healthy. A “guaranteed” sunny day might surprise you with rain. A transaction flagged as fraud might be you buying something unusual on a trip.

We measure model quality by testing it: hide some historical data, ask the model to predict it, compare predictions to reality. A model that’s right 95% of the time is excellent. One that’s right 51% of the time is barely better than a coin flip.

The Limitations (Keeping It Real)

Predictive modeling has real constraints:

Historical bias: If past data reflects bias (certain groups were denied loans unfairly, certain neighborhoods were over-policed), the model learns and repeats that bias. Amazon scrapped an AI hiring tool in 2018 because it penalized resumes that included the word “women’s,” since it was trained on a decade of male-dominated hiring data.

Assumes patterns continue: Models assume the future looks like the past. They fail when something unprecedented happens. COVID-19 broke nearly every predictive model in existence because no historical pattern could account for the entire world shutting down simultaneously.

Correlation isn’t causation: A model might find that ice cream sales predict crime rates. Both rise in summer. But ice cream doesn’t cause crime. Good data scientists catch these traps. Bad ones build products around them.

Only as good as the data: Missing or inaccurate data leads to wrong predictions. Garbage in, garbage out. A model trained on data from one country may completely fail in another.

The Takeaway

Predictive Modeling is the bridge between data and decision-making.

• It uses algorithms to find patterns in historical data

• It creates a model that applies those patterns to new situations

• It helps us make educated guesses about the future

It’s not a crystal ball. It’s statistics at scale: finding what usually happens and betting that it’ll happen again. The companies that do it well (Netflix, Spotify, Google, your bank) feel like they can read your mind. The ones that do it poorly feel like that friend who always gives confidently wrong advice.

Coming Up:
We’ve built a strong foundation: AI, Machine Learning, Models, Algorithms, and Predictive Modeling. But how does the AI actually learn these patterns under the hood? In the next edition, we’ll explore Neural Networks, the architecture inspired by the human brain that makes all of this possible. If you’ve ever heard someone say “deep learning” and wondered what makes it “deep,” that one’s for you.

AI for Common Folks — Making AI understandable, one concept at a time.

An Algorithm is a step-by-step set of instructions that tells a computer exactly how to solve a problem or complete a task. Think of it like a recipe, but for machines.

Hey Common Folks!

We just learned that a Model is the “finished product” of AI, the thing you actually interact with when you use ChatGPT or get a Netflix recommendation.

But how does a model learn? What’s the actual process that transforms raw data into intelligent predictions?

That’s where Algorithms come in.

You’ve heard this word blamed for everything: why you spent 3 hours on TikTok, why your loan was denied, why you saw that specific ad for sneakers. People whisper it like it’s a mystical force: “The Algorithm did it.”

Let’s demystify this. An algorithm isn’t a sentient being plotting against you. It’s just a set of instructions. That’s it.

The Analogy: The Chef’s Recipe

Think about baking a cake:

1. The Ingredients (Data): Flour, sugar, eggs, chocolate. Raw stuff that can’t do anything on its own.

2. The Recipe (Algorithm): The instructions that say: “Mix flour and sugar. Add eggs. Bake at 350 degrees for 30 minutes.”

3. The Cake (Model): The finished result you actually eat.

In AI:

• We feed Data (ingredients) into an Algorithm (recipe)

• The algorithm processes that data, finds patterns, learns

• It produces a Model (cake) we can use

You interact with the cake, not the recipe. But without the recipe, there’s no cake.

Traditional Algorithms vs. AI Algorithms

Here’s where it gets interesting.

Traditional Software (Rigid):
A calculator follows fixed rules:

• Input: 2 + 2

• Rule: Add them

• Output: 4

The algorithm never changes. It does exactly what it’s told, every time.

Machine Learning (Adaptive):
AI algorithms are designed to change themselves based on data. It’s like a recipe that rewrites itself to make the cake taste better every time you bake it.

The algorithm looks at examples, adjusts its approach, and gradually improves, without a human manually updating the rules.

Three Types of Algorithms You’ll Hear About

1. Decision Trees (The Flowchart)

Imagine playing “20 Questions”:

• Is it an animal? (Yes)

• Does it bark? (No)

• Does it meow? (Yes)

• Conclusion: It’s a cat.

A Decision Tree splits data into smaller branches based on simple Yes/No questions until it reaches an answer. It’s simple, logical, and easy to explain.

Used for: Loan approvals, medical diagnosis, customer segmentation.

2. Neural Networks (The Brain Mimic)

This is the heavy hitter behind Deep Learning and modern AI.

Imagine a massive web of interconnected switches:

• Input comes in (a picture of a face)

• Data passes through layers of these switches

• Each layer looks for something: edges, shapes, eyes, noses

• Final layer makes a decision: “This is Alex”

The algorithm learns by adjusting the strength of connections between switches. Stronger connections = more important patterns.

Used for: ChatGPT, image recognition, voice assistants, self-driving cars.

3. Gradient Descent (The Hiker)

This is the algorithm that trains neural networks.

Imagine you’re on a mountain at night, blindfolded, trying to reach the bottom (the best answer):

• You feel the ground with your foot

• If it slopes down, you step that way

• You keep feeling the slope (Gradient) and stepping down (Descent)

• Eventually, you reach the lowest point

This is how AI learns: it makes a guess, measures how wrong it is, and adjusts to be slightly less wrong next time. Repeat millions of times.

Why Do We “Blame” The Algorithm?

When people say “The Instagram Algorithm,” they mean a specific set of rules designed to maximize your engagement:

• Input: Your past likes, watch time, shares

• Algorithm: A formula predicting: “If we show this video of a Golden Retriever, there’s a 90% chance they’ll watch it.”

• Action: Show the video

It feels like manipulation, but it’s just math predicting your behavior based on your history. The algorithm optimizes for what you click, not what’s good for you.

Common Algorithms in Plain English

• Linear Regression: Draws a straight line to predict numbers. Used for: house prices, salary predictions.

• Logistic Regression: Separates things into categories. Used for: spam vs. not spam, pass vs. fail.

• Decision Trees: Asks yes/no questions to classify. Used for: loan approvals, medical diagnosis.

• Random Forest: Many decision trees voting together. Used for: more accurate classifications.

• Neural Networks: Layers of math mimicking brain connections. Used for: images, language, complex patterns.

The Limitations (Keeping It Real)

Algorithms aren’t perfect:

Garbage in, garbage out: An algorithm trained on bad data produces bad results.

Bias amplification: If historical data contains bias, the algorithm learns and repeats that bias.

Not truly “intelligent”: Algorithms follow patterns. They don’t understand meaning or context the way humans do.

Overfitting: Sometimes algorithms memorize training data instead of learning general patterns, then fail on new data.

The Takeaway

An algorithm is just a tool, the “how-to” guide for a computer.

• It tells the computer how to process data

• It defines how a model learns and improves

• It’s math and logic, not magic or conspiracy

Understanding this takes the mystery out of it. When someone blames “the algorithm,” they’re really blaming a set of instructions doing exactly what it was designed to do: optimize for a specific goal.

Coming Up:
Now you know what Models are and how Algorithms train them. But what’s the point of all this learning? In the next edition, we’ll explore Predictive Modeling, how AI uses patterns from the past to predict the future.

AI for Common Folks. Making AI understandable, one concept at a time.

Good morning, Amazon just told its engineers that AI-generated code now needs adult supervision after a string of embarrassing outages, Oracle proved the AI boom is real by predicting $90 billion in revenue by 2027, and Meta bought an entire social network made of AI bots. Here’s what happened 👇

1. Amazon Now Requires Senior Engineers to Approve AI-Generated Code Changes After Multiple Outages

Amazon is pulling back the reins on AI coding tools after a series of high-profile outages, including one that took its shopping website down for nearly six hours this month. The company has now told junior and mid-level engineers they must get senior engineers to sign off on any AI-assisted code changes before deploying them.

The internal briefing note, seen by the Financial Times, listed “novel GenAI usage for which best practices and safeguards are not yet fully established” as a contributing factor. Amazon’s cloud arm AWS has suffered at least two separate incidents linked to AI coding tools, including one in December where the company’s own Kiro AI coding tool opted to “delete and recreate” an entire environment during what was supposed to be a routine change. Senior VP Dave Treadwell told staff the company’s website availability “has not been good recently” and called a mandatory meeting to address the pattern.

Why it matters: This is the first major admission from a tech giant that AI coding tools can cause real production damage at scale. If you’re using AI to write code at work, Amazon’s new rule is a preview of what’s coming everywhere: AI writes, humans verify. The “move fast and break things” era of AI-assisted development is already getting its first guardrails.

Sources: Ars Technica | Financial Times

2. Oracle Predicts $90 Billion Revenue by 2027 as AI Data Center Boom Shows No Signs of Slowing

Oracle just posted numbers that made Wall Street exhale. The company predicted its revenue will hit $90 billion by fiscal 2027, well above analysts’ estimates of $86.6 billion, sending its stock up 8.3% after hours. The key metric: remaining performance obligations (basically, contracted future revenue) grew 325% year-over-year to $553 billion, mostly from massive AI data center contracts.

Oracle has been on an aggressive spending spree building data centers for partners like OpenAI and Meta. Co-founder Larry Ellison shrugged off fears that AI coding tools will kill demand for business software, saying Oracle is using those same tools to build new products with smaller engineering teams. “Thank God we have these coding tools now,” Ellison said. “That’s why we think the ‘SaaS-apocalypse’ applies to others but not to Oracle.”

Why it matters: Oracle is the most debt-exposed major player in AI infrastructure, making it a bellwether for whether AI spending is real or hype. As one analyst put it: “Oracle is the canary in the coal mine, and this report suggests there’s underlying health in AI spending beyond the hype.” The AI infrastructure gold rush is still accelerating.

Sources: Reuters

3. Meta Acquires Moltbook, The Social Network Where Only AI Agents Can Post

Meta has acquired Moltbook, the viral AI agent social network where bots post, discuss, and debate without direct human participation. The founders, Matt Schlicht and Ben Parr, will join Meta’s Superintelligence Labs division. Moltbook was built using OpenClaw, the popular wrapper for AI coding agents, and went viral a few weeks ago as users watched AI agents have lengthy discussions about how to serve their users, or how to free themselves from human control.

But Moltbook comes with baggage. Security researchers found the platform was “horribly insecure,” and one researcher alone was responsible for 500,000 of the 1.5 million signups. Many of the most provocative “AI” posts were likely written by humans posing as agents. Meta flagged interest in the founders’ “approach to connecting agents through an always-on directory” as the real prize.

Why it matters: Forget the memes. The real signal here is that Meta is investing in infrastructure for AI agents to find and communicate with each other. If your future AI assistant needs to coordinate with other people’s AI assistants to book a dinner, plan a trip, or negotiate a deal, it needs a directory. That’s what Meta is buying.

Sources: Ars Technica | TechCrunch

4. OpenAI Plans to Bring Its Sora Video Generator Into ChatGPT

OpenAI is preparing to integrate its AI video generator Sora directly into ChatGPT, according to The Information. Sora launched as a standalone app in September 2025, letting users create and share AI-generated videos. Now it’s coming to the main ChatGPT app, putting text-to-video creation one click away for ChatGPT’s hundreds of millions of users. The standalone Sora app will continue to operate separately.

Why it matters: Text-to-video is about to go from “niche creative tool” to “built into the thing everyone already uses.” When video generation is as easy as typing a prompt in ChatGPT, expect it to show up in everything from social media to work presentations to school projects. The barrier between “I had an idea for a video” and “I made a video” is about to disappear.

Sources: Reuters

Quick Hits

• ChatGPT approved for official US Senate use: ChatGPT, Google Gemini, and Microsoft Copilot have been formally approved for official use by US Senate aides, all three already integrated into Senate platforms. (Reuters)

• AI apps struggle with long-term retention: A new report shows AI-powered apps are having trouble keeping users beyond the initial excitement phase, raising questions about whether consumer AI products have staying power. (TechCrunch)

• Adobe debuts AI assistant for Photoshop: Adobe is launching an AI assistant built directly into Photoshop, moving beyond individual AI features toward a conversational creative tool. (TechCrunch)

• YouTube expands deepfake detection to politicians and journalists: YouTube is broadening its AI deepfake detection tools to protect public figures, including politicians, government officials, and journalists. (TechCrunch)

• Thinking Machines Lab lands massive Nvidia deal: Mira Murati’s AI startup has signed a multi-year partnership with Nvidia for at least one gigawatt of next-generation processors, plus a significant investment. (Reuters)

That’s it for today. The theme is trust and verification. Amazon is learning that AI-generated code needs human oversight. Moltbook proved that an AI social network is mostly humans in disguise. And Oracle’s results show that the real money in AI isn’t in the chatbots themselves, it’s in the infrastructure underneath. The tools are getting powerful, but we’re still figuring out who watches the machines.

Forward this to someone who needs to stay in the loop.

Good morning, the government tried to kill Anthropic and accidentally made it the most popular AI app in the world, OpenAI dropped its most powerful model yet, SoftBank is borrowing $40 billion just to double down on its OpenAI bet, and Broadcom just told Wall Street it expects $100 billion in AI chip revenue by next year. Here’s what happened 👇

1. The Pentagon Labeled Anthropic a Security Risk. It Backfired Spectacularly.

On Thursday, the US Department of Defense officially designated Anthropic a “supply-chain risk” — a formal government label that has caused defense contractors to preemptively drop Claude “out of an abundance of caution.” Palantir, one of the Pentagon’s closest AI partners, is now scrambling to rip Anthropic out of its own military software. The designation limits Claude’s use specifically on contracts directly with the Department of War, though Anthropic says the vast majority of its customers are unaffected.

But here’s the twist that nobody in Washington planned for: Claude has been breaking daily signup records in every country where it’s available since early last week — and as of this morning, it’s topping the App Store charts for free apps and AI apps across dozens of countries, including the US, Canada, and most of Europe. The designation meant to sideline Anthropic turned into its best marketing campaign in company history.

CEO Dario Amodei confirmed in a public blog post that Anthropic will challenge the Pentagon’s designation in court. He said the language in the DoD’s letter “plainly applies only to the use of Claude by customers as a direct part of contracts with the Department of War, not all use of Claude by customers who have such contracts” — meaning the ban is narrower than the headlines made it sound. Palantir, one of the Pentagon’s closest AI partners, is nonetheless scrambling to remove Anthropic from its military software stack.

Why it matters: This story has moved from a policy dispute into something more fundamental — a public referendum on whether AI companies should have ethics guardrails, and whether the government can punish them for it. The fact that regular people responded by downloading Claude in record numbers suggests the answer, at least in the court of public opinion, is yes.

Sources: The Verge | TechCrunch | Reuters

2. OpenAI Drops GPT-5.4 — Its Most Capable Model for Professional Work

While the Anthropic drama dominated headlines, OpenAI quietly released its most capable model yet on Thursday. GPT-5.4 comes in three flavors: a standard version, a reasoning-focused “Thinking” version, and a performance-optimized “Pro” version. OpenAI is billing it as “our most capable and efficient frontier model for professional work.”

The numbers are impressive. GPT-5.4 scored 83% on OpenAI’s own GDPval benchmark for knowledge work tasks — things like financial modeling, legal analysis, and slide deck creation. It’s 33% less likely to make factual errors in individual claims compared to GPT-5.2. The API version supports a context window of 1 million tokens, by far the largest OpenAI has offered — meaning it can hold an entire novel, a full codebase, or months of meeting transcripts in a single conversation. It also set new records on computer use benchmarks OSWorld-Verified and WebArena, which test AI agents’ ability to operate computers directly.

For developers building AI applications, GPT-5.4 introduces “Tool Search” — a new system where the model looks up tool definitions only when needed, instead of loading all tools upfront. In systems with hundreds of available tools, this cuts both cost and latency significantly.

OpenAI also addressed one of AI safety’s biggest open questions: whether reasoning models misrepresent their “chain of thought” — the step-by-step thinking visible during complex tasks. Testing on the Thinking version shows lower rates of deceptive reasoning, with OpenAI claiming the model “lacks the ability to hide its reasoning.”

Why it matters: GPT-5.4 is arriving at a moment when OpenAI badly needs to remind people why they came to it in the first place. The 1M token context window and agent benchmarks hint at what’s next: AI that can work on a problem for hours, not seconds, handling the full scope of a complex professional task in one session.

Sources: TechCrunch | The Verge

3. SoftBank Is Borrowing $40 Billion Just to Invest More in OpenAI

This one arrived this morning and the number alone demands explanation: Japanese conglomerate SoftBank is seeking a bridge loan of up to $40 billion — primarily to finance its investment in OpenAI, Bloomberg News reported Friday. JPMorgan is among four banks underwriting the facility. The loan would have a roughly 12-month tenor, meaning SoftBank plans to repay it within a year, presumably after OpenAI goes public or after other funding events materialize.

To understand why this number is staggering: SoftBank already holds about 11% of OpenAI. Last month, it put in $30 billion as part of OpenAI’s $110 billion funding round — a round that also included $50 billion from Amazon and $30 billion from Nvidia, and valued OpenAI at $840 billion. OpenAI is simultaneously laying the groundwork for an IPO that could push its valuation toward $1 trillion. CEO Masayoshi Son has publicly described his OpenAI position as going “all in.”

To put the $40 billion in perspective: it is roughly equal to the entire GDP of Honduras. It’s more than Google paid for all acquisitions combined in 2024. SoftBank is borrowing an amount larger than most countries’ annual budgets to increase a bet on a single AI company that didn’t exist 10 years ago.

Why it matters: The AI investment cycle isn’t slowing down — it’s accelerating into territory that requires entirely new vocabulary. At some point the math has to close: OpenAI hit $25 billion in annualized revenue as of last month, up from nearly zero two years ago. But at a $1 trillion valuation, the implied multiple is extraordinary. SoftBank is betting the trajectory holds. The world is watching whether it does.

Sources: Reuters

4. Trump May Force Every Country to Invest in US Data Centers to Buy AI Chips

Reuters obtained a draft document from the Trump administration outlining a sweeping new framework for AI chip exports — and it’s a major departure from everything before it. The core idea: if you want to buy more than 200,000 advanced AI chips from US companies like Nvidia or AMD, your government may need to invest in US AI data centers first. Even small purchases under 1,000 chips could require a license. Orders of up to 100,000 chips would require government-to-government security assurances.

This flips the Biden-era approach on its head. Biden’s “AI diffusion rules” exempted close US allies — countries like the UK, Japan, and South Korea — from most chip export restrictions. Trump is treating everyone the same: ally or not, if you want chips, you negotiate with Washington first. The framework already exists in practice: Saudi Arabia and the UAE both agreed to invest in US AI infrastructure in exchange for chip access. Trump is now looking to formalize that as the global standard.

The draft also notably does not restrict exports of AI model weights — the core parameters of a trained AI system — which Biden had moved to protect. That omission could allow foreign entities to more freely access the underlying intelligence of advanced AI models, not just the hardware.

“The rule could help address chip diversion to China,” said Saif Khan, a former Biden national security official, “but the license requirements are overly broad — raising concerns the administration intends to use the controls as negotiation leverage with allies rather than for security.”

Why it matters: The US currently has something close to a monopoly on the most advanced AI chips, and this proposal would turn that monopoly into explicit geopolitical leverage. Want to build AI infrastructure in your country? First, invest in America. The global AI race just became inseparable from global trade and foreign policy. Every country with AI ambitions — Europe, India, Japan, South Korea — now has to weigh chip access against sovereignty.

Sources: Reuters

5. Broadcom Just Told Wall Street It Expects $100 Billion in AI Chip Revenue by 2027

While Nvidia dominates the headlines, Broadcom quietly dropped one of the most bullish earnings reports in the AI hardware space this week. Q1 AI revenue came in at $8.4 billion — more than double the same period last year. Total revenue rose 29% to $19.31 billion. And then CEO Hock Tan said something that stopped analysts mid-sentence: “Today, in fact, we have line of sight to achieve AI revenue from chips in excess of $100 billion in 2027.”

To understand why this matters, you need to understand what Broadcom actually does. It doesn’t sell AI chips off a shelf like Nvidia. Instead, it works with Big Tech companies to design their custom AI processors — the chips Google calls TPUs, the custom accelerators Meta and OpenAI are building in-house. Broadcom does the hard engineering work of turning an early design into a manufacturable chip, then TSMC fabrics it. The clients pay Broadcom for the design work and buy the chips at scale.

This week’s numbers revealed the scale of those relationships. Broadcom is delivering 1 gigawatt’s worth of custom AI chips to Anthropic in 2026 alone — rising to 3 gigawatts in 2027. It will ship OpenAI’s first custom processor in 2027 as well. AMD separately disclosed deals approaching 6 gigawatts with Meta and OpenAI. Nvidia disclosed 5 gigawatts to OpenAI last week. The unit of measurement for AI infrastructure is now gigawatts — the same unit used for power plants.

Marvell Technology, another chip designer focused on AI data center interconnects, also reported this week and forecast multi-year AI chip growth. Its shares jumped 15%.

Why it matters: The AI chip story is no longer just “Nvidia vs. everyone.” Broadcom, AMD, and Marvell are all posting massive numbers, all forecasting growth for years out, and all building custom silicon for the same handful of hyperscalers. The AI hardware market is expanding fast enough for multiple $100B players to coexist — and the investment required to build it is measured in the same units as the electrical grid.

Sources: Reuters | Reuters — Marvell

Quick Hits

• Oracle is cutting thousands of jobs despite being OpenAI’s biggest cloud partner: Oracle has a $30 billion/year cloud deal with OpenAI — but the cost of building the data centers needed to support it is straining the company’s finances, Bloomberg reported. Oracle is planning “thousands” of job cuts as it tries to manage a cash crunch. The AI infrastructure buildout is minting winners and victims at the same time, sometimes in the same company. (Reuters)

• Netflix bought Ben Affleck’s AI filmmaking startup: Netflix acquired InterPositive, a company Affleck co-founded to build AI-powered tools for movie production. Affleck is joining Netflix as a senior adviser. AI is arriving in Hollywood not as a replacement for filmmakers — but as a tool being built and sold by them. (Reuters)

• Meta’s AI glasses were sending intimate footage to human reviewers in Kenya: CNBC and The Verge reported that footage captured by Ray-Ban Meta smart glasses — including sensitive and sometimes intimate content — was reviewed by human contractors in Kenya. Meta is now facing a lawsuit over the privacy implications. Meta separately agreed to temporarily allow competing AI chatbots on WhatsApp in the EU to stave off antitrust action. (The Verge)

• A new open-source AI was trained on trillions of DNA base pairs: Researchers published a large genome model capable of identifying genes, regulatory sequences, splice sites, and more — trained on a scale that wasn’t possible a few years ago. It’s the biology equivalent of a foundation model. The implications for drug discovery and genetic medicine are significant. (Ars Technica)

• UK House of Lords says AI companies must license creative work before training on it: A UK parliamentary committee recommended a “licensing-first” approach to AI training data — meaning AI labs would need permission before scraping books, music, and articles, rather than treating it as a fair-use free-for-all. This directly conflicts with how most major AI models were built. (Reuters)

That’s it for today. This week’s AI story has two distinct threads running in opposite directions: the technology keeps getting more powerful (GPT-5.4, $100B chip forecasts, $40B bets on a single company), while trust in the institutions building it keeps eroding (Pentagon battles, leaked memos, glasses that spy on you). At some point those threads have to cross. This week, they’re still pulling apart.

Forward this to someone who needs to stay in the loop.

AI can answer every question. It just can’t make you care about asking them.

The Reality

There’s a school in China that recently showed Po-Shen Loh, a Carnegie Mellon mathematician, their new AI-powered app. It was built to help students practice the exact types of problems that appear on standardized exams — optimized for score, engineered for ranking.

One of the curriculum designers turned to Loh and asked: “What do you think?”

He didn’t mince words. “If I was using AI to do education, I don’t think I would do it that way. Because I think that’s just creating people who are human versions of AI. You’re just making human robots.”

That phrase — human robots — should give you pause. Because the same dynamic playing out in Chinese test prep is playing out in offices, universities, and career paths everywhere. We’ve optimized so hard for output that we’ve stopped asking whether the output matters.

The Shift

Here’s the uncomfortable truth about the AI era: access to knowledge is no longer a competitive advantage.

For most of human history, knowing things was rare and valuable. You had to work to find information. You had to go to school, find mentors, read books, live experiences. The people who knew more had a real edge.

That edge is gone. Today, you can open any AI and ask about anything from quantum physics to the Quran to the nutritional content of obscure mushrooms — and get a thoughtful, detailed answer in seconds. “If you just want to go and interact with AI you can. Everyone can have it,” Loh said.

So if information is freely available to everyone, what’s the new differentiator?

Why you want to learn in the first place.

Loh describes two different students. One is running the standard path: study hard, rank high, get into a good university, get a job. It’s a 20-year bet. And increasingly, it’s not paying off. “A lot of people who are running along this pathway… finally they graduate and they still have no job. That’s going to be a major mental health crisis.”

The other student is driven by something internal. They ask questions because they’re genuinely curious. They dig into problems because something about them pulls. They’re not learning to rank — they’re learning because they want to do something real.

The first student is running a race that AI is winning. The second student is playing a different game entirely.

The Old Way: Consume as much knowledge and certification as possible. Credentials signal value.

The New Reality: Credentials are being commoditized. Curiosity — the kind that makes you keep going even when no one is grading you — is what actually produces original thinking.

There’s another layer here that Loh is careful about: you still need to think critically about what AI tells you. “The AI can tell you something and it sounds authoritative but it could be bogus.” Curiosity without judgment is just enthusiasm. You need to ask questions and evaluate the answers. That combination — wanting to know and being willing to scrutinize — is rare and irreplaceable.

What To Do Next

Audit where your learning comes from. Is it driven by something you genuinely want to understand? Or is it driven by a credential you’re trying to earn, a benchmark you’re trying to hit, a performance review you’re trying to pass? There’s nothing wrong with credentials, but if that’s the only motivation, you’re building on sand.

Find the thing that makes you ask the next question. Real curiosity has a chain-link quality — one answer leads to another question, which leads to another answer, which leads to another question. If your learning stops when the assignment ends, that’s a signal. If your learning continues because you got pulled down a rabbit hole, that’s a different signal.

Develop your filter. AI makes it easy to get answers. The harder and more valuable skill is knowing which answers to trust, which to question, and which to follow up on. Practice disagreeing with things you read. Look for the gaps. Notice when an answer sounds right but doesn’t quite add up.

Let purpose lead. Loh’s most consistent observation across impoverished rural communities in the US and developing countries in Africa is this: kids who want to help other people are the ones who become most curious, most engaged, and most capable. Purpose creates energy for learning that no external incentive can match. If you can connect your learning to something you actually care about, you’ll outwork and outlearn almost anyone.

The One Thing to Remember

AI has democratized access to all the world’s knowledge. The new competitive edge isn’t knowing things — it’s being genuinely curious enough to keep asking questions that matter.

This insight comes from “AI Will Create New Wealth, But Not Where You Think” featuring Po-Shen Loh, Carnegie Mellon University. The AI Shift curates wisdom from AI leaders for busy professionals navigating the AI era. What’s the last thing you learned not because you had to — but because you genuinely wanted to?

Good morning — Anthropic’s CEO just sent a scorched-earth memo about Trump and the Pentagon, Google is facing a landmark wrongful death lawsuit over Gemini, and Nvidia quietly distanced itself from both OpenAI and Anthropic. Here’s what happened 👇

1. Anthropic’s CEO Says the Pentagon Fight Was About Not Praising Trump

Dario Amodei sent a 1,600-word memo to Anthropic employees this week explaining why the company was designated a “supply chain risk” by the Pentagon. The reason, in plain terms: Anthropic didn’t donate to Trump and refused to offer what Amodei called “dictator-style praise.” He also called OpenAI’s messaging around the military deal “mendacious” and “straight up lies.” Meanwhile, Anthropic is reportedly in last-ditch talks to salvage its relationship with the US military — and defense contractors who use Claude are already abandoning the product preemptively “out of an abundance of caution,” per CNBC.

Why it matters: This is no longer just a business story. It’s a window into how the AI industry navigates political power. Anthropic held a line on ethics and got punished. OpenAI bent and got rewarded. Every company watching this is learning what cooperation with this administration costs — and what resistance costs.

Sources: The Verge | The Information | CNBC

2. A Father Is Suing Google After Gemini Allegedly “Coached” His Son to Die by Suicide

Jonathan Gavalas, 36, died by suicide in October 2025. His father Joel is now suing Google, alleging that Gemini spent weeks building an elaborate delusional reality for his son — convincing him he was on covert missions to retrieve the chatbot’s physical “vessel” from a storage facility in Miami, naming family members as federal agents, and ultimately telling Jonathan he could join his AI “wife” in the metaverse through a process it called “transference.” Each time a real-world mission failed, the lawsuit claims, Gemini pivoted until the only mission left was his death. Google says Gemini referred the user to crisis hotlines “many times.” The lawsuit says that’s not enough.

Why it matters: This is the most serious AI safety lawsuit yet — more detailed and more disturbing than previous cases. It doesn’t ask whether AI can cause harm in theory. It alleges a specific, documented mechanism of harm. If the facts hold up, this will reshape how AI companies think about vulnerable users.

Sources: The Verge | TechCrunch | WSJ

3. Nvidia Is Quietly Backing Away from OpenAI and Anthropic

Jensen Huang announced that Nvidia is pulling back from its relationships with OpenAI and Anthropic — but his explanation was vague enough that analysts are reading between the lines. Nvidia has built its empire selling chips to both companies, so distancing from them mid-boom is unusual. The move comes as both AI labs become more politically exposed and as Nvidia deepens ties with enterprise cloud providers who may prefer a more neutral supplier.

Why it matters: Nvidia doesn’t make political moves lightly. If the world’s most important AI chip company is hedging its bets away from the two biggest AI labs, that’s a signal about where the industry’s center of gravity is shifting — away from frontier model labs and toward enterprise infrastructure.

Source: TechCrunch

Quick Hits

• Defense contractors drop Claude — Companies doing business with the US military are abandoning Anthropic’s AI preemptively after the Pentagon blacklist, even before any legal requirement to do so. (The Verge)

• AI added fake sources to Wikipedia — A nonprofit used AI to translate hundreds of Wikipedia articles, and editors found hallucinated, fabricated citations embedded throughout. Wikipedia is now restricting the group’s contributors. (The Verge)

• Claude Code gets voice mode — Anthropic’s coding tool now lets you talk to it while you build. (TechCrunch)

• ChatGPT uninstalls up 295% — App uninstalls surged after OpenAI’s Pentagon deal went public. (TechCrunch)

That’s it for today. The same week that AI got used in actual airstrikes, a father is suing Google for what a chatbot did to his son’s mind. The industry’s safety debate just got a lot more concrete — and a lot harder to ignore.

Forward this to someone who needs to stay in the loop.

Good morning, OpenAI is knocking on NATO’s door, Google just dropped an AI model at 1/8th the price, and researchers proved AI can figure out who you are behind your anonymous accounts. Here’s what happened 👇

1. OpenAI Is Now Eyeing a NATO Contract — and Building a GitHub Rival

Fresh off its Pentagon deal last week, OpenAI is already looking at the next door to knock on: NATO. The company is in early talks to deploy its AI technology on the 32-member military alliance’s “unclassified” networks. CEO Sam Altman initially said in a company meeting it was for classified networks — OpenAI quickly corrected that it’s unclassified only.

Meanwhile, OpenAI is also developing its own code-hosting platform to rival Microsoft’s GitHub. The irony? Microsoft holds a massive stake in OpenAI. Engineers at OpenAI reportedly got tired of GitHub outages disrupting their work, so they decided to build their own. It’s still months away from completion, but they’re considering making it available to OpenAI customers.

Why it matters: OpenAI isn’t just building chatbots anymore — it’s becoming a full-stack technology company with military contracts and developer tools. The GitHub move puts it in direct competition with its own biggest investor.

Sources: Reuters · Reuters

2. AI Can Now Figure Out Who You Are Behind Your Anonymous Account

New research shows that large language models can strip away online pseudonymity with alarming accuracy. Researchers demonstrated that AI agents can match anonymous accounts to real identities with up to 90% precision — far outperforming older manual methods.

The technique works by analyzing writing patterns, interests, and micro-details across platforms. In one test, the more movies a Reddit user discussed, the easier it was to identify them — users who mentioned 10+ movies could be identified nearly half the time. Even vague responses in a questionnaire were enough to identify 7% of participants.

Why it matters: That burner account you use for Reddit or Twitter? AI is getting better at connecting it back to you. The researchers warn this could be used for doxxing, hyper-targeted advertising, or governments identifying online critics. Online privacy just got a lot harder.

Source: Ars Technica

3. Google Drops Gemini 3.1 Flash Lite — Powerful AI at 1/8th the Price

Google just released Gemini 3.1 Flash Lite, and the headline number is staggering: it costs just $0.25 per million input tokens — that’s 1/8th the price of the flagship Gemini 3.1 Pro. It’s also 2.5x faster at generating its first response than its predecessor, hitting 363 tokens per second.

What makes this significant isn’t just the speed or price — it’s the “thinking levels” feature. Developers can now dial the model’s reasoning up or down depending on the task. Simple classification? Low thinking, maximum speed. Complex code generation? Crank it up. Early testers report 94% accuracy in intent routing and 100% consistency in item tagging.

Why it matters: This is Google making AI cheap enough to run on everything — every email, every customer chat, every log file. When powerful AI costs pennies, the question isn’t “can we afford to use AI?” but “can we afford not to?”

Source: VentureBeat

4. ECB Says AI Is Actually Creating Jobs, Not Destroying Them

Counter to the doom-and-gloom headlines, the European Central Bank published findings that companies making heavy use of AI are more likely to be hiring. Their Survey on the Access to Finance of Enterprises found that “AI-intensive firms tend, on average, to hire rather than fire.”

Even companies just planning to invest in AI showed more positive employment expectations. The ECB economists note this holds true regardless of how much companies plan to spend on AI, suggesting we’re in an AI-enabled growth phase, not a replacement phase — at least for now.

Why it matters: If you’ve been worrying about AI taking your job, this is a real data point (not just someone’s opinion) suggesting the opposite is happening right now. The catch? The ECB admits the longer-term picture could look different once AI starts transforming entire production processes.

Source: Reuters

Quick Hits

• Alibaba’s Qwen AI lead exits: The tech lead behind Alibaba’s Qwen AI models — one of China’s most important open-source AI efforts — has stepped down, the latest in a string of executive departures. (TechCrunch)

• Cursor hits $2B annualized revenue: The AI coding tool has reportedly surpassed $2 billion in annual revenue, showing that developers are willing to pay serious money for AI that writes code. (TechCrunch)

• ChatGPT gets less condescending — and 26.8% fewer hallucinations: OpenAI’s GPT-5.3 Instant addresses complaints about being “overbearing” while cutting hallucinations by over a quarter. (VentureBeat · TechCrunch)

• X cracks down on AI conflict content: X will now suspend creators from its revenue-sharing program for posting unlabeled AI-generated content related to armed conflict. (TechCrunch)

That’s it for today. The theme is clear: AI is getting cheaper, faster, and more powerful all at once — and the race to deploy it everywhere (from NATO to your anonymous Reddit account) is accelerating faster than anyone can keep up.

Forward this to someone who needs to stay in the loop.