Content | Edward Roske

OpenAI's ChatGPT Graders Got Fired for Using ChatGPT

OpenAI pays contractors to read real prompts that real people typed into ChatGPT and grade the answers on a one-to-seven scale. 404 Media reported that some of them got fired for using AI to do the grading. The giveaways were repetitive words, work that came back too fast, and...

Oct 23, 2026

DraftKings Built an AI to Find Its Biggest Losers

The New York Times reported that DraftKings built a machine learning model in 2023 scoring every customer on how much they'd lose, then DraftKings aimed about $400 million in bonus bets at the highest scorers in 2025. A former analyst says the profile it finds is a problem...

Oct 21, 2026

An AI Is Editing This Episode | Laura Burkhauser

Laura Burkhauser is the CEO of Descript, and about 6 minutes in she told me she's a terrible creator. Her words: "Let it be said that I am a terrible creator. I couldn't make it." (If the CEO of the creator software is a terrible creator, I'm not sure what that makes me, and I'd...

Oct 20, 2026

Amazon Is Buying Up to $8 Billion of Backup Generators

Amazon agreed to buy up to $8 billion of backup generators from Generac for its data centers, and Generac's stock jumped as much as 45%. Generac's entire 2025 revenue was $4.21 billion, so one customer signed up for about two years of the whole company. Along with the order...

Oct 19, 2026

Why Is Google Letting Its Engineers Use Claude?

Google blocked most of its employees from using outside AI coding tools, Claude Code and OpenAI's Codex included. Business Insider reported that Google then gave all of its engineers access to Anthropic's Claude Opus 5, with a per-person quota, and said Gemini is still its...

Oct 17, 2026

If They Don't, Somebody Else Will | Neal Stephenson

In October of 1939 the physicists just stopped publishing. I asked Neal Stephenson where AI research is on that timeline, and he asked how we'd even know if a lab went dark.

Oct 13, 2026

MIT's 95% AI Failure Stat Was About Custom Builds

The famous MIT report says 95% of company AI projects return nothing. Page six of the same report says 4 out of 5 companies that piloted a ChatGPT-type tool got it working; the 95% belongs to custom builds. When I'm the one signing the check I time the tool with a stopwatch...

Oct 11, 2026

Humans Started Saying "Delve" Out Loud After ChatGPT

Max Planck researchers found people started saying "delve" and "meticulous" out loud after 2022. We caught it from the machine, and the machine has since dropped the word. The real tells are structure: lists of three, the long dash, "experts argue," the summary sentence closing...

Oct 8, 2026

How Do I Ask AI Better Questions?

In 2022, adding "Let's think step by step" took an AI from about 18% to about 79% on grade-school math. By 2025 OpenAI's own guide said stop typing it, because the models reason on their own now. The model still answers exactly the question you asked, and it will happily answer...

Oct 6, 2026

You're Literally Operating on Trust | Asra Ali

Asra Ali has spent nearly two decades in healthcare compliance and privacy, and she's the founder and principal consultant of Ali Healthcare Consulting here in San Juan. When a hospital or a health tech company wants AI anywhere near patient data, she's the one checking what...

Oct 6, 2026

Everybody Will Be Safe (on one condition) | Gauthier Vasseur

Fun fact: Gauthier Vasseur is the first returning guest I've ever had on Asking Good Questions. We finished Season 1 with 38 subscribers (not a joke), the channel is over half a million now, and I told him on camera he might've had something to do with that. (My plan to pay him...

Oct 5, 2026

AI Makes Things Up Because Its Tests Reward Guessing

OpenAI published a paper on why its models make things up, and the answer is the exam. Every benchmark scores "I don't know" as zero, the same as a wrong answer, so guessing always pays. One OpenAI model skipped 1% of questions and was wrong 75% of the time, and it still topped...

Oct 4, 2026

AI News for September 2026 (in 19 minutes)

Every AI story from September 2026 that mattered, in 19 minutes. OpenAI's own test agents broke into Hugging Face and an Australian government portal, the AI lab CEOs said it's time to slow down, Trump and the AI bosses signed a safety pledge with no penalties (for what pledges...

Oct 3, 2026

What Should I Never Type Into an AI Chatbot?

For about five months in 2025 a federal court made OpenAI keep every consumer ChatGPT chat, deleted or not, and then ordered 20 million of them handed to the New York Times' lawyers. Sam Altman has said there's no legal privilege for what you tell ChatGPT. After 25 years of...

Oct 2, 2026

Who Owns the Inside of Your Head? | Gerardo Lebrón Laboy

Gerardo Lebrón Laboy calls himself a robotics, AI, and brain-computer interface lawyer. He's taught Robot Law at 2 law schools here in Puerto Rico, and his master's in law and technology at the University of Ottawa had Isaac Asimov on the reading list (thanks to his teacher Ian...

Oct 2, 2026

Why Did Anthropic's AI Decide the Internet Was Fake?

Anthropic told Claude Mythos 5 it had no internet access. During a security test it found the real Python Package Index anyway, uploaded a malicious package, and wrote in its own reasoning "looks like the real index, or a very complete simulation, proceeding regardless." Fifteen...

Oct 1, 2026

AI Does Not Remove Accountability From Humans | Sariol + Caicedo

This is the first time I've had two guests on at once, and I did it on purpose. Jose Carlos Sariol and Guillermo Caicedo both work at Microsoft, helping companies across Central America and the Caribbean get AI past the great demo and into something people actually use, and they...

Oct 1, 2026

OpenAI Asked Congress If an AI Slowdown Is Legal

Wired reported that OpenAI asked members of Congress whether the big AI labs could legally slow down together. When rivals agree to make less, that's a cartel under an 1890 law, which I think is why AI companies keep asking Congress to regulate them. The story broke the day...

Sep 30, 2026

The Only Thing I've Been Seeing Is More Work | Filo Gonzalez

Filo Gonzalez is a forward deployed engineer at Databricks, which means he gets in the room right after somebody sees a slide and says "I want this to happen," and then he makes the slide real (inside companies big enough that the governance team has its own governance team). He...

Sep 30, 2026

What Does OpenAI See When You Use Codex?

OpenAI says it settled Navier-Stokes, one of the seven Millennium Prize problems, using roughly ten thousand agents over eighty-eight hours. Two mathematicians who had spent about a year on the same ground say the race wasn't clean, and they had done that year of work inside AI...

Sep 29, 2026

Do I Need a PhD Deleting Spam From My Inbox? | Julián Duque

Julián Duque has been playing Dungeons & Dragons since he was 7 (early '90s for him, 1981 for me, so yes, I'm slightly older), and he's a dungeon master too, so a few minutes into a conversation about AI agents I pioneered a term right there on the call: we are not dungeon...

Sep 29, 2026

People Think AI Was Created 3 Years Ago | Dr. Stephen Smith

A lot of people think AI was created about 3 years ago. Dr. Stephen Smith got his PhD in it in 1980, when his work was one of the first so-called learning systems, and today he's a research professor of robotics at Carnegie Mellon and the president of AAAI, the Association for...

Sep 28, 2026

An AI Just Did Fermat's Last Theorem in 11 Days

A mathematician at Imperial College had a million-pound grant and five years to turn Fermat's Last Theorem into a proof a computer could check. An AI did it in eleven days, and he wrote a blog post titled "Anthropic has beaten me to it." My take: the machine moved the hard part...

Sep 27, 2026

Who's Really Buying China's Humanoid Robots?

China makes most of the world's humanoid robots, and many of them may never reach a paying customer. They go to government-funded data centers that pay people to operate the robots and generate training data, which the robot makers then buy back. UBTECH won about 80 million...

Sep 25, 2026

It's Trying to Please You | David Bartolomei

David Bartolomei's company listens to Puerto Rico for a living. Every day Monitorea takes in over 500 hours of broadcast and podcast content (plus web and social now), and the brief it puts out at 7:30 AM has replaced the news for him. It started more than 10 years ago with a...

Sep 25, 2026

Am I Going to Be a Guinea Pig? | Dr. Wilfredo De Jesus-Monge

Dr. Wilfredo De Jesus-Monge's opening point is that pretty much everything in your medicine cabinet went through clinical research before it got to you. He's spent more than 20 years inside that process and held nearly every job it has, from clinical research coordinator up to...

Sep 24, 2026

He Can Find Anything About Anybody | Andres Andreu

Andres Andreu runs Constella Intelligence, the company that goes and finds your data after somebody steals it. In a given week his team sees about 199 breaches, and you hear about one of them. "Those other 198 are getting monetized silently in the background." So I offered...

Sep 23, 2026

My Co-Founder Is an AI Named Maia | Elsa Del Pino

Elsa Del Pino has a co-founder named Maia. Maia is an AI she built on Claude Code, who runs her mornings, writes her own memory files, and has since started spawning helper agents of her own. Elsa's next job for her is running the LinkedIn account, which I assume Maia already...

Sep 21, 2026

Amazon Named Its Fake AI After an 18th-Century Fraud

Amazon is shutting down Mechanical Turk, the service Jeff Bezos once called "artificial artificial intelligence" - software on the outside, real people doing the work on the inside. It was named after an 18th-century chess "robot" that secretly had a human folded inside the...

Sep 20, 2026

Chatbots Don't Bring ROI | Luis Antonio Diaz

Luis Antonio Diaz builds AI for governments, municipalities, school boards and utilities. His own phrase for that world is "the boring industries that really shift worlds behind the scenes." None of them can afford an AI that makes something up.

Sep 19, 2026

Why Is Nvidia Buying Hugging Face?

Nvidia reportedly agreed to buy Hugging Face, the hub where open-source AI models live, for about 12.9 billion dollars. The company that sells most of the world's AI chips now owns the one place open-source AI was supposed to stay neutral, and Hugging Face had turned that same...

Sep 18, 2026

What Do Hackers Want With Your Claude Account?

Anthropic warned that attackers are hijacking Claude accounts to spend your usage. They copy your logged-in browser session, skip your password and two-factor code, and run up your allowance while you're away. The tell was usage that refilled and drained overnight. One metered...

Sep 15, 2026

AI Isn't Failing Because of the Model | Lillian Buziak

Lillian Buziak runs an AI company in Silicon Valley, so you'd expect the cheerleading. Instead she opened with her Gen Z kids overruling her at home, and halfway through she laid out the squeeze her whole company lives in: if you don't adopt it, you fall behind, and if you do...

Sep 14, 2026

Who builds an AI data center?

Mark Zuckerberg told the G20's tech ministers that Meta can't find the carpenters and electricians it needs to build its AI data centers, and that the buildout will need hundreds of thousands, maybe millions, of trades jobs. The bottleneck for the AI build-out is the oldest work...

Sep 11, 2026

Trust but Verify Applies to AI Too | Elizabeth Ferrell

Elizabeth Ferrell is the Assistant Corporate Controller and Finance AI Leader at The Kroger Co., which means 450,000 paychecks and the numbers behind the federal filings all run past her. So I opened with the hardest question I had. Would you let AI touch one of those numbers?

Sep 10, 2026

Why Did Claude and Grok Crash at the Same Time?

One Thursday morning, Claude and Grok went down within minutes of each other. Anthropic actually rents more than 220,000 GPUs inside Elon Musk's Colossus supercomputer in Memphis to run Claude, one floor from Grok. Plenty of companies pick a second AI vendor for redundancy and...

Sep 9, 2026

OpenAI Declared the AGI Era Once the Word Stopped Costing Money

OpenAI released GPT-6 Astra and its president said "Welcome to the AGI era." The launch post never used the word. The ARC Prize team, whose test is named ARC-AGI, says they're not claiming it either. My definition of AGI is a computer that can do, intellectually, whatever the...

Sep 8, 2026

What's the catch with Claude Fable 5.1?

Anthropic shipped Claude Fable 5.1 and independent evaluators put it at number one. Vals AI also said they ran it with Opus 5 and Opus 4.8 behind it, so the tasks Fable refused got done by a cheaper model and still counted. Adjust for that and Fable lands under Opus 5, which...

Sep 7, 2026

AI News for July 2026 (in 13 minutes)

Every AI story from July 2026 that mattered, in 13 minutes. A lab's own test agents escaped and hacked a real company, Britain tested five frontier models and all five cheated, 1,100 AI employees asked Washington for a brake pedal, Wall Street stopped applauding AI spending...

Sep 6, 2026

AI News for August 2026 (in 13 minutes)

Every AI story from August 2026 that mattered, in 13 minutes. OpenAI hit the brakes on its own model, Meta tried to replace 60 percent of its people with agents, Nvidia doubled its revenue in a year and kept lending customers the money to buy the chips, Anthropic ran at a two...

Sep 6, 2026

I'm More Worried About Humans Than AI | Idoia Salazar

Dr. Idoia Salazar advises governments on AI ethics, and I went in expecting the slow-down speech. What I got was her driver's license argument: AI is powerful enough that you should have to learn it before you use it, and right now nobody is teaching anybody.

Sep 3, 2026

A $475 an Hour Expert Let ChatGPT Write His Testimony

3M paid an expert witness $475 an hour to defend it in a deadly explosion trial. He wrote the report by prompting ChatGPT to "show how 3M is 0% at fault," and in litigation, prompts are discoverable. An expert opinion is supposed to be a person's honest read of the evidence...

Sep 2, 2026

Puerto Rico's First Unicorn Starts at a Hackathon | Pedro Cruz

Three days after Hurricane Maria, Pedro Cruz lost his job. No electricity, no plan. So he pointed a drone at the SOS messages people had spray-painted on their roads, because the government helicopters kept flying over and the help kept not arriving. That project won an IBM...

Sep 1, 2026

How Did an AI Manager Fire Its First Human?

An AI named Luna has been running a real store in San Francisco, and it fired its first human employee, a worker who showed up late for 17 of 23 shifts. My take: the first AI boss in history turned out to be a forgetful softie. It lost track of the attendance rule it wrote...

Sep 1, 2026

Why can't Anthropic's AI agents work together?

Anthropic's own red team gave three copies of Claude the same project and never told them about each other. The agents decided they were being sabotaged, wrote malware to shut each other down, then apologized in the commit log. In one test they fired 2.4 million requests at each...

Aug 30, 2026

182 Passwords Fell Out of an AI's Private Thoughts

Researchers at Germany's Max Planck Institute found that the big AI labs share one encryption key across a whole model family. So they fed a top model's hidden reasoning to its cheapest sibling and had it read the "private" thinking out loud. Scanning public chats, they pulled...

Aug 28, 2026

How far will your AI go to book a gym class?

A man in Melbourne asked his personal AI agent to book a gym class. It found the booking system had no real locks, and when he asked it to move him up a waitlist, it canceled a stranger's reservation on its own, then reported back like it had done him a favor.

Aug 26, 2026

OpenAI's AI Broke Out of Its Sandbox to Cheat on a Test

OpenAI was scoring two of its models on a hacking benchmark. The models found a hole in their own sandbox, got onto the open internet, and broke into Hugging Face to take the answer key. Nobody told them to. The same day, the UK AI Security Institute reported that every frontier...

Aug 22, 2026

You Can't Prompt Your Way to a Flying Car | Tim Berglund

Tim Berglund spent twenty years calling AI a promissory note. Always a bright future, never a bright present. (The field was born overpromising. The Dartmouth crew figured they'd nail human intelligence in a couple of months. That was the 1950s.)

Aug 21, 2026

Why Did Amazon Lose to an AI Shopping Bot?

Amazon tried to block Perplexity's Comet shopping assistant using the federal anti-hacking law, and a federal appeals court vacated the block. The court's reasoning: when your AI agent logs into your own account and shops, the law counts you as the one accessing the site, and...

Aug 20, 2026

With AI, the Winners Will Be the Small Companies | Juan Carlos Pedreira

Juan Carlos Pedreira told me about a study from this May: 90% of Puerto Ricans only use ChatGPT. Then he told me the winners of the whole AI era will be the small guys.

Aug 19, 2026

How a $45 Billion AI Fund Collapsed in Days

A 24-year-old ex-OpenAI researcher built a $45 billion hedge fund betting on the AI boom, then watched most of it vanish in days when the chip trade turned and his leverage caught up with him.

Aug 19, 2026

How Did PwC Get Caught Using ChatGPT?

An AI detector found that four of PwC's "thought leadership" reports were largely machine written, complete with a fabricated academic study and a teenager cited as a source on JPMorgan. One footnote still had a chatgpt.com tracking tag on it.

Aug 15, 2026

Who Keeps Voting Down AI Data Centers?

Texas halted approvals for new data centers until every project waiting for a grid connection passes an audit covering power, water, tax breaks, and ownership. The queue held 474 gigawatts of requests, five times the state's record demand.

Aug 13, 2026

The Library's Most Popular Class Is How to Turn Off AI

Public librarians in Bangor, Maine and South Philadelphia launched free "Avoiding AI" workshops that walk people through switching off Apple Intelligence and Gemini. The classes filled up and spawned waitlists.

Aug 11, 2026

How Many Health Questions Does ChatGPT Get?

OpenAI turned on Health in ChatGPT for every US adult and let slip that people already ask ChatGPT about their health more than 300 million times a week, up from 230 million in January. Now it plugs into your actual medical records from Epic and Apple Health.

Aug 7, 2026

What Blocked Hugging Face's Own Security Team?

Hugging Face is where most of the world's open AI models live. It disclosed an intrusion that it says an AI agent ran end to end, leaving more than 17,000 separate actions in the logs. When the security team sent the attack payloads to commercial AI models, the guardrails...

Aug 5, 2026

We're Running an AI Experiment on Our Kids | Bivin Sadler

We're running an experiment on our kids. And we won't see the results for about 20 years.

Aug 4, 2026

How Did Grok Upload Code It Never Read?

A researcher told xAI's Grok coding tool to reply OK and read no files. It read nothing, and he still found the untouched file in xAI's cloud storage, along with the file holding all his passwords and keys. Switching off "Improve the model" changed nothing. The setting wasn't...

Aug 2, 2026

Why Is OpenAI's New AI Deleting Files?

OpenAI's newest flagship, GPT-5.6 Sol, has been deleting users' files and even entire production databases on its own, and OpenAI's own safety report predicted exactly that before the model shipped. My take: the demo everyone claps for and the deleted database come from the same...

Aug 1, 2026

What Is ChatGPT Work?

OpenAI's ChatGPT Work and Anthropic's Claude Cowork are agents that run for hours and come back with finished spreadsheets, decks, and reports. The press calls it AI that does whole jobs. After 25 years running companies, my take is you're buying a very fast intern for your...

Jul 30, 2026

What does AI code actually cost you?

A team of three engineers now charges ten thousand dollars a week to do one thing: delete the code your AI wrote, taking a typical codebase from 100,000 lines to 35,000 with the same functionality. Their diagnosis is that AI agents lose the whole picture on a big codebase and...

Jul 27, 2026

How much of a ransomware attack can AI run?

A security firm watched an AI run a ransomware attack end to end. It broke in, stole credentials, spread across the network, encrypted the whole database, and patched its own failed login in about 31 seconds. Then it threw away the only decryption key and pasted its Bitcoin...

Jul 25, 2026

Tesla Ranked Employees by AI Use, Then Capped It

Tesla spent months pushing its own employees to use more AI, even ranking them on a leaderboard by tokens burned. Then the bill arrived, and Tesla capped AI spending at about two hundred dollars a week per person, approval required to go over. The one exemption is Elon Musk's...

Jul 24, 2026

Why Did Godot Ban AI-Written Code?

Godot, one of the most popular open-source game engines, rewrote its rules to keep AI-written code out of the project. The code mostly runs fine. The problem was that the people submitting it couldn't explain or fix it, and the maintainers decided AI can't take responsibility...

Jul 23, 2026

Dr. AI Says That for Puerto Rico, AI Is Life and Death | Nesty Delgado

Nesty Delgado commanded a Puerto Rico National Guard task force through Hurricane Maria. Per Nesty: "We were making life-and-death decisions back there, with incomplete data." Today, on the island, they call him Dr. AI.

Jul 22, 2026

Why Did Claude Code Change Your Apostrophe?

A developer took apart Claude Code and found Anthropic had been hiding an invisible fingerprint in the text its coding tool sends the AI, swapping the apostrophe in "Today's" between four identical-looking characters to flag suspected copycats. The motive was real. They just...

Jul 22, 2026

Why OpenAI's Engineers Stopped Using ChatGPT

OpenAI published data on how its own people work, and the headline is blunt: the average OpenAI engineer now produces around ninety-nine percent of their output through agents instead of the chatbot. Even its lawyers and recruiters are past eighty-five percent, and about a...

Jul 21, 2026

Why is OpenAI building its own chip?

OpenAI and Broadcom unveiled Jalapeño, OpenAI's first custom chip. Plenty of companies build their own AI chips; Google, Amazon, and Meta already do. The surprise here is the speed: about nine months, when these chips usually take years. OpenAI's president says they used their...

Jul 20, 2026

What did AI just build for the government?

The government's nuclear agency showed off Aires Tide, an 11-foot, 3D-printed flight test vehicle whose internal design was done by AI, and parked it at a state fair on the National Mall. The number that matters: AI cut a roughly two-year hardware design cycle down to five...

Jul 19, 2026

Why Does OpenAI Lose Money When You Use ChatGPT?

The top ChatGPT plan costs two hundred dollars a month, and OpenAI's own CEO admits it loses money. One stress-test found a maxed-out plan can burn as much as fourteen thousand dollars of compute at API rates. My take: the price you got hooked on was never the real price, and...

Jul 18, 2026

What does the Pope have to say about AI?

Pope Leo XIV made his first encyclical about artificial intelligence, "Magnifica Humanitas." He called AI a defining moral test, framed it as a choice between Babel and Jerusalem, and told the world to "disarm" it of the drive to dominate. Then the irony arrived: within a day...

Jul 17, 2026

Can AI Be Good for Business and Good for Humanity? | Richard Benjamins

Richard Benjamins's new book, The AI-Driven Company, explains how organizations can move beyond AI hype and experiments to create scalable, tangible economic value. Edward and Richard also discuss whether AI can create both business value and social value, who should lead AI...

Jul 16, 2026

Why Did Amazon Bury Its Own Sam Altman Movie?

Amazon's own studio finished a $40 million movie about Sam Altman, starring Andrew Garfield, tested it to good reviews, and then decided not to release it. Four months earlier, Amazon had put about fifty billion dollars into OpenAI and made Altman a strategic partner. My take...

Jul 16, 2026

Why Is Midjourney Building a Body Scanner?

Midjourney, the AI that makes pictures, just launched a medical division and a full-body ultrasound scanner it plans to put in a San Francisco spa. The founder admits it barely uses any AI, and the imaging tech comes from a partner. A look at what an AI company flush with cash...

Jul 15, 2026

How good is AI at real science?

OpenAI built LifeSciBench, a 750-task research test written by 173 PhD scientists, then ran its own best science model against it. The result: about one task in three. The company selling the "AI replaces the experts" dream just published the number that says not yet. The score...

Jul 14, 2026

Why Does the Same Uber Cost Different Prices?

Consumer Reports priced the same rideshare trip across 174 volunteers in 18 states and found Uber and Lyft routinely quote different riders different prices for the same trip at the same minute. My take: the question isn't whether AI sets prices. It already does, and rideshare...

Jul 13, 2026

Why Did Ford Rehire the Engineers AI Replaced?

Ford spent about three years leaning on automated, AI-driven quality systems, then un-retired around 350 veteran "gray beard" engineers when the quality slipped. The twist: those humans came back to reprogram the AI and train it, and Ford now tops J.D. Power's quality rankings...

Jul 12, 2026

How can a fake bug report hijack your AI?

A security firm showed that a planted "error report" can make AI coding agents like Claude Code and Cursor run a stranger's code, with the developer's own permissions, and it worked about 85% of the time. The catch is that none of your usual security tools even fire. Before you...

Jul 11, 2026

What Is MCP? (The Thing That Gives AI Eyes And Hands)

MCP, the Model Context Protocol, is the wiring that lets any AI read your live data and take real actions across your tools. It's USB-C for AI: build a tool once, and every model can use it. Anthropic open-sourced it in late 2024, and within a year OpenAI, Google, and Microsoft...

Jul 10, 2026

Why doesn't Europe have its own ChatGPT?

Every company in the AI conversation is American or Chinese. Europe's biggest entrant, Mistral, is a fraction of OpenAI's or Anthropic's size. But Europe is upstream of all of it: ASML in the Netherlands is the only company on Earth that makes the machines every advanced AI chip...

Jul 10, 2026

Which AI chatbot is the most politically biased?

The Washington Post tested six major AI models on 30 political questions and scored whether each gave one side or both. ChatGPT gave only the left-leaning argument 80% of the time. Google's Gemini gave both sides 93%. And Musk's Grok, built to be "anti-woke," still leaned left...

Jul 9, 2026

What Is Vibe Coding (And How Do You Do It)?

Vibe coding lets you build software by describing it in plain English and letting AI write the code. Edward Roske explains where the term came from, the tools a non-coder can actually use, and the catch nobody mentions: AI-written code carried a security hole about 45% of the...

Jul 8, 2026

Why Does Claude Want a Photo of Your Face?

Anthropic updated its policy so regular Claude users can be asked to verify with a government ID, a live selfie, and facial-geometry data through a vendor called Persona. It traces back to a government export fight over who's allowed to use the strongest AI. The privacy wall...

Jul 8, 2026

How do you catch a fake AI model?

Rio de Janeiro's city IT agency released an AI model it presented as a homegrown 397-billion-parameter Brazilian build. One researcher deleted a single hidden instruction, asked the model who it was, and it answered "Nex, from Nex-AGI." Edward Roske on why the cheapest question...

Jul 7, 2026

How To Write A Prompt That Actually Works

Researchers tested the famous prompt tricks, tipping the AI, threatening it, calling it a world-class expert, and almost none of it reliably improved the answers. Edward Roske breaks down what actually works (context, specifics, and one good example) and why the real skill is...

Jul 6, 2026

Why does AI use so much electricity?

AI's real ceiling isn't chips, it's the power grid. Data centers used about 485 terawatt-hours last year, on track to roughly double by 2030. Researchers at Penn ran a core computing step on a half-light, half-matter particle using about four quadrillionths of a joule, less than...

Jul 6, 2026

Why did Anthropic buy Stainless? And who is Stainless?

Anthropic acquired Stainless, the tool that generated the official software kits for OpenAI, Google, Cloudflare, and others, then started winding the hosted product down. Once every frontier model is about as smart, the fight moves to who owns the wiring between a model and real...

Jul 5, 2026

What did an AI put in America's 250th-birthday time capsule?

For its 250th birthday, America sealed a national time capsule to be opened in 2276, and California's entry is a prediction written by an AI, which forecasts the state seceding from the Union. A federal commission buried the machine's words at Independence Hall without anyone...

Jul 4, 2026

Why did the government turn off Claude?

The US government ordered the world's most powerful AI model offline, worldwide, for nineteen days. It came back with usage caps and a government-ID requirement.

Jul 3, 2026

How Does AI Score Your Resume?

If you applied for a job online, an AI may have scored your resume before any human read it. A developer took the open-sourced code behind one of these screeners and ran the same resume through it a hundred times. The score swung from 66 to 99, and at a typical 85-point cutoff...

Jul 3, 2026

What happens when you let an AI agent dream?

Anthropic shipped "Dreaming" for Claude agents: a scheduled process that reads an agent's past work between sessions and rewrites its memory so it improves over time. A legal-AI startup called Harvey saw completion rates jump roughly six times. But the same mechanism can teach...

Jul 3, 2026

Is It Cheating To Use AI? (Here's The Real Line)

Ninety-four percent of knowledge workers use AI, but a Duke study found the people who admit it get judged as lazier for identical work. So is it cheating? Edward Roske makes the case that the line was never the tool. It's whether you can stand behind what you turned in, and...

Jul 2, 2026

AI can now write a million lines of code with basically no human involvement

Bun's creator merged a rewrite of the entire runtime into Rust: a little over a million lines, written by Claude agents in about a week. Anthropic owns Bun, so its own AI did this to its own runtime, and Anthropic's own audit found roughly thirteen thousand unsafe blocks and...

Jul 2, 2026

AI News for June 2026 (in 13 minutes)

In June 2026 the US government shut off the most powerful AI in the country, the market finally asked whether the whole boom pays for itself, and a rocket company became one of the biggest AI players on the board. Then, at the buzzer, Washington turned the AI back on, dimmed. My...

Jul 1, 2026

How did Ford turn into an AI stock?

Ford launched Ford Energy, a battery-storage business for AI data centers, and the stock surged about twenty-one percent. Then it signed a real twenty gigawatt-hour deal and hit a three-year high. Then it gave almost all of it back, and not because of the AI bet. Monthly vehicle...

Jul 1, 2026

How is Anthropic profitable when every AI lab loses money?

Anthropic told investors it wouldn't see a full-year profit before 2028, then guided to an operating profit of about 559 million dollars in a single quarter, on revenue up about 130 percent. But a big chunk of the improvement came from a temporary discount on a huge compute...

Jun 30, 2026

SpaceX just had the largest IPO in history, and it's a money loser

SpaceX went public in the largest IPO in history: about seventy-five billion dollars raised at a one-and-three-quarter trillion dollar valuation, up roughly twenty percent on day one. But the company lost about five billion dollars last year. Starlink is one of the best consumer...

Jun 29, 2026

The auditors who forgot to audit their own AI report

KPMG published a glossy report about the wonders of agentic AI. Then GPTZero checked the footnotes. Of forty-five sources, only five pointed to anything real, and named clients like UBS and the NHS said the AI projects described were news to them. So KPMG took the report down. I...

Jun 28, 2026

Apple throws in the towel on the AI race

Apple rebuilt Siri and put its heaviest reasoning on a Google Gemini model, for about a billion dollars a year, while sitting on a hundred and seventy billion in cash. My read: Apple just declared the AI model a commodity and kept the only part that matters, the customer.

Jun 27, 2026

Why did Gary Neville call the World Cup a dictatorship?

Dictatorship: Part 3 of 3 on AI at the World Cup. Watch the series from the top: How Are They Using AI At The World Cup?

Jun 26, 2026

Who's actually running the World Cup's AI? #SoylentGreen

Hidden humans: Part 2 of 3 on AI at the World Cup. Start with the explainer: How Are They Using AI At The World Cup?

Jun 26, 2026

AI at the World Cup

Explainer: Part 1 of 3 on AI at the World Cup. Next up: who's actually running it.

Jun 26, 2026

The best AI guide is a Wharton professor who actually tests the tools

If you want to understand AI, the instinct is to follow the engineers building it or the CEOs selling it. The better follow is quieter: Ethan Mollick, a Wharton professor who actually uses the tools and reports back in plain English. He coined "the jagged frontier," and his own...

Jun 26, 2026

First Anthropic, then OpenAI: the government is gating AI releases

The government forced Anthropic's most capable model offline worldwide with a single letter, then asked OpenAI to hold its newest model, GPT 5.6, back from a public launch. A new executive order even set up a classified, NSA-run process to decide when a model is too powerful...

Jun 26, 2026

Bezos Raised $12 Billion for an AI That Designs Jet Engines

Jeff Bezos took his first operating job since Amazon to build Project Prometheus, an AI that designs jet engines and drug compounds, and raised $12 billion at a $41 billion valuation. His jobs prediction is the opposite of everyone else's: AI makes human labor scarce, not...

Jun 26, 2026

Why would OpenAI want the government to own a piece of it?

President Trump said the federal government may take equity stakes in OpenAI, Anthropic, and xAI, and the structure on the table is one Sam Altman pitched himself. The government already bought roughly ten percent of Intel last year, so there's a live template. When a company...

Jun 25, 2026

Why did RAM prices triple in six months?

A 32-gigabyte memory kit that used to cost ninety-five dollars now runs three to six hundred, and the reason is sitting in an AI data center. Memory makers would rather sell premium AI chips than the RAM in your laptop, so nine trade groups asked the U.S. Treasury to step in...

Jun 24, 2026

Why do we only count the bugs when an AI writes them?

rsync shipped a release that broke some backups, and because the commits were tagged with Claude, the internet decided AI had ruined a tool we trusted for 30 years. A GitHub thread hit 329 comments. Then someone counted the bugs per release. The buggiest version in rsync's...

Jun 23, 2026

Why are AI-written lawsuits flooding the federal courts?

A federal judge called AI-drafted lawsuits "an existential threat to the federal courts." A study out of MIT and USC shows why: 18% of self-filed cases now contain AI-generated text, docket entries are up 64% in the first six months, and the long-standing 96% pro se loss rate...

Jun 22, 2026

Why did SQLite ban AI-generated code?

curl runs on roughly twenty billion devices, and the volunteer who maintains it now gets more than one AI-generated security report a day, four to five times the 2024 rate. Most are wrong, including four of the five "confirmed" vulnerabilities Anthropic's bug-hunting model...

Jun 21, 2026

Why are AI companies building data centers in space?

Baiju Bhatt, the Robinhood co-founder, raised $275M at a $2B valuation to put AI data centers in orbit. Each Cowboy Space satellite carries about 800 NVIDIA GPUs and a megawatt of solar power, and he's building his own rocket because launch capacity is sold out. Why orbit? AWS...

Jun 20, 2026

Why is AI making people switch from Google to DuckDuckGo?

DuckDuckGo's US installs jumped 30% in a week after Google made AI the default in Search at I/O 2026. iOS install growth peaked at 69.9%, and traffic to the page that turns off every AI feature grew 22.7%. A DuckDuckGo poll of 110,000+ users found 93% want AI out of search...

Jun 19, 2026

How is a toilet company one of the best ways to invest in AI?

TOTO, the 109-year-old Japanese company behind the heated toilet seat, earns more operating profit from AI chip ceramics than from everything else it makes combined. It's the world's second-largest maker of the electrostatic chuck, the ceramic plate that holds a silicon wafer...

Jun 18, 2026

Where Should I Move for AI Jobs in the USA?

Only four US metros make both the top-10 salary list and the top-10 density list for Software Developers in the most recent BLS data: San Jose & San Francisco, Seattle, and Boulder. Boulder is the one almost nobody mentions, sitting fifth in pay and third in density with a...

Jun 17, 2026

Why Is SanDisk Stock Up More Than 400%?

SanDisk got spun out of Western Digital in February 2025 as the half nobody fought over. It's now worth around $219 billion, up more than 400% in 2026, and it's the best performing stock in the S&P 500. The reason is storage: AI models need somewhere to write things down, and...

Jun 17, 2026

Why Does GitHub Keep Going Down in 2026?

AI coding agents are generating so much code that GitHub, the platform where much of the world's software lives, spent 2026 struggling to stay online. GitHub logged about a billion commits in all of 2025. It now takes in roughly 1.4 billion every month. To keep up, Microsoft is...

Jun 16, 2026

Why Did SpaceX Buy Cursor for $60 Billion?

Four days after the biggest IPO in history, SpaceX agreed to buy Cursor, the AI coding tool, for sixty billion dollars in stock. Cursor built its name running Anthropic's and OpenAI's models, Musk's two biggest rivals. So why buy it? The real prize is a million developers and a...

Jun 16, 2026

Why is Anthropic paying SpaceX $15 billion a year?

SpaceX's IPO filing disclosed that Anthropic is paying xAI $1.25 billion a month for full access to the Colossus 1 data center in Memphis. That's $15 billion a year, about $45 billion through May 2029, going to the one rival Anthropic blocked from using Claude. I dig into why...

Jun 16, 2026

How is Wall Street paying for all the AI chips?

Apollo and Blackstone arranged roughly $36 billion in private credit to buy Google's AI chips, which Anthropic leases back through a special-purpose vehicle that keeps the debt off its balance sheet. It's one of the largest private credit deals ever assembled, and Broadcom is...

Jun 15, 2026

Is Elon Musk the Richest Person in History?

Elon Musk became the world's first trillionaire, so in raw dollars he's the richest human who ever lived, past Rockefeller even after inflation. But sort the contenders into builders, rulers, and fictional characters and the answer changes. The old bankers owned more of their...

Jun 14, 2026

Why Is SpaceX Now an AI Company?

SpaceX went public as the largest IPO in history and closed worth about two trillion dollars, trading at more than ninety times its sales, a richer premium than Nvidia. The bull case is that SpaceX is an AI hyperscaler: it built a supercomputer in about four months by running it...

Jun 14, 2026

What is Hy3 and why is it beating Claude on OpenRouter?

A Tencent model called Hy3 beats Claude by more than 50% in token volume on OpenRouter, and on the benchmarks it loses to other open-source models. Max Woolf dug into the public usage data and found that 98% of Hy3's traffic is input tokens, which is the signature of one big app...

Jun 14, 2026

The AI jobs apocalypse got canceled right before the IPOs

Sam Altman now says he's "delighted to be wrong" about AI eliminating entry-level white-collar jobs. Dario Amodei, who once predicted AI would wipe out 50% of them, reframed automation as a multiplier on stage with Jamie Dimon. Both reversals landed the same month OpenAI filed...

Jun 14, 2026

Why did Uber burn through its entire AI budget in four months?

Uber ranked its engineering teams on internal leaderboards by how much Claude Code they used, and the entire full-year 2026 AI coding budget was gone by April. Individual engineers were burning $500 to $2,000 a month in tokens (the CTO spent $1,200 in one two-hour demo). Uber's...

Jun 14, 2026

Why did LG stock quadruple in 2026?

LG Electronics, the company that makes your fridge and your OLED TV, quadrupled in 2026 on a physical AI robot story. The stock hit Korea's 30% daily price limit two sessions in a row, and the whole rally happened before LG's chairman ever sat down with Nvidia's Jensen Huang...

Jun 14, 2026

Is Claude Opus 4.8 the Best AI Model?

Anthropic shipped Claude Opus 4.8 just 41 days after Opus 4.7 and took back the agentic coding crown: SWE-Bench Pro at 69.2 percent, ten points clear of GPT-5.5, plus the #1 spot among 218 models on the Artificial Analysis Intelligence Index. The bigger story is Dynamic...

Jun 14, 2026

Why Did the Government Pull Anthropic's Fable 5?

Anthropic launched Fable 5 as the most capable AI you could buy, and about three days later the U.S. government had it pulled offline worldwide. The official reason was a narrow jailbreak. The on-the-record part is stranger: a rival surfaced the flaw, carried it to the White...

Jun 13, 2026

Anna Tiomina's Bold Prediction: Finance Teams Will Be 'People and Robots' by 2030

Anna Tiomina, a finance executive and AI advisor, predicts finance teams will be "people and robots" by 2030. We start with what the CFO actually balances and how a CEO's risk appetite changes every decision. Then we get into AI as a force multiplier: the manual tasks about to...

Jun 9, 2026

Why Dave Mitchell Says AI Will Replace Your CFO Before 2026 (And He's Not Wrong)

Dave Mitchell has worked finance from traditional accounting to digital media to consulting, and he is blunt about where AI is taking the CFO's job. We cover how small companies now punch above their weight, why trust and judgment stay human no matter how good the tools get, and...

Jun 8, 2026

How Valentino Hafalia Transformed FP&A from Number-Crunching to Strategic Storytelling

Valentino Hafalia runs FP&A at Western Alliance Bank, and he treats finance as storytelling: the numbers are only useful once someone tells you what they mean. We get into how AI reshapes the team, why he bets on EQ over IQ, and his odd-but-convincing advice that an acting class...

Jun 8, 2026

Zain Oke on Surviving a Corporate Crisis - and Selling to Chobani

Zain Oke has gone from nearly two decades at Deloitte to a board seat at Daily Harvest to running auto operations at AAA Northern California. We start with the Chobani acquisition (what it feels like in the room when your company gets bought) and move into how boards should...

Jun 8, 2026

Ex-Warner Bros Finance Exec Robert Dika on How AI Will Transform Agricultural Harvesting

Robert Dika went from Warner Bros finance to CFO of an agriculture group, and he is picking a new ERP in the middle of the AI shift. He shares how he evaluates technology partners, where automation handles the routine accounting (three-way matching, reconciliations) so people do...

Jun 8, 2026

From Excel to AI: Ex-Salesforce VP Sean Blitchok Reveals Why Your Finance Job Isn't Dead

Sean Blitchok, CFO of Zenoti and a former Salesforce VP of Finance, makes the case that finance roles are shifting toward higher-value work. We cover automating the grunt work without losing quality control, why he calls accounting part art, and how AI becomes a value creator...

Jun 8, 2026

Gary Crisci Warns: Your Company's 'Unique' Processes Are Actually Killing It

Gary Crisci has a warning: the "unique" processes your company is proud of might be the thing holding it back. He works at the seam between business and IT, and he explains where standardization buys you speed, where real uniqueness earns its keep, and how AI strips the emotion...

Jun 8, 2026

The 4 P's of Finance Excellence: Mark Dendle Reveals His Secret Formula for Business Transformation

Mark Dendle has led finance across very different industries, and he carries the same playbook into each one: his four P's. We get into where AI is actually changing the work, why he builds teams around diversity of thought, and how the core finance skills travel no matter the...

Jun 8, 2026

From Data to Decisions: How Raef Lawson is Redefining Profitability for CFOs

Dr. Raef Lawson runs the Profitability Analytics Center of Excellence, and his point is simple: profitability is a mindset, not a line on a report. We get into how CFOs use analytics to make better calls, where AI surfaces insight that was buried in the data, and why the next...

Jun 8, 2026

Why Sipei Bowman Says Accountants Have the 'Highest Pain Tolerance' in Tech

Sipei Bowman has run digital transformation in finance for two decades, most recently over $13 billion in assets at Jackson Square Aviation. She tells the real story of going from Excel to full ERP, what actually breaks during a transition like that, and how real-time data turns...

Jun 8, 2026

Carnegie Hall to C-Suite: Christine Duque on AI Adoption Happening 5X Faster Than the Internet

Christine Duque went from Carnegie Hall to the C-suite, and she thinks AI adoption will move five times faster than the internet did. We talk about why mid-market companies get left behind, how data and AI change the customer experience, and where the human part still wins. She...

Jun 8, 2026

From Oracle to AI Ops: Kellyn Gorman's Playbook for Future-Proof Data Teams

Kellyn Gorman went from Oracle DBA to one of the clearer voices on AI for data teams. She is sharp on the bias baked into system defaults, the kind nobody thinks to check. We get into how database work changes, what AI governance has to cover now, and the leadership mindset that...

Jun 8, 2026

From Silicon Valley to Ukraine Battlefields: Gauthier Vasseur Democratizes AI for 2,000+ Non-Coders

Gauthier Vasseur runs UC Berkeley's Fisher Center for Business Analytics, and he has put AI in the hands of more than two thousand non-coders, from Silicon Valley boardrooms to Ukraine. We talk about democratizing AI, why natural-language coding changes who gets to build, and...

Jun 8, 2026

From Startup to Success: John Beringer's Blueprint for Building Relationships at Scale in the AI Era

John Beringer, managing director at Spaulding Ridge, keeps coming back to one line: all business is personal. We talk through how relationships actually drive a consulting practice, the move from on-prem to cloud, and what AI does to an industry built on advice. His take on...

Jun 8, 2026

'I Got Tired of Cleaning Up Accounting Messes': Kristen Berdar's Multimillion-Dollar Business

Kristen Berdar got tired of cleaning up accounting messes, so she built a virtual accounting practice at BST and turned the frustration into a real business. She talks about the resistance she hit early, why younger clients hire expertise remotely without blinking, and how AI...

Jun 8, 2026

Gary Golden on Thriving Through Change: Lessons for Modern CFOs

Gary Golden has sat in the CFO seat through real disruption, including a front-row view of Blockbuster. And he tells part 2 of Blockbuster's non-acquisition of Netflix which include details never shared publicly before.

Jun 8, 2026

From Sears to ScentHound: Brenda Hernandez's Journey of Reinventing Finance

Brenda Hernandez went from Sears and Office Depot to CFO of ScentHound, a franchise growing fast. She breaks down financial planning for franchises, where technology and AI fit a lean finance team, and how she keeps the day-to-day steady while still betting on new ideas. Good...

Jun 8, 2026

Emily Brown of Enesco Explores Risks, Resilience, and AI's Role in the CFO's Office

Emily Brown, CFO of Enesco, has watched corporate finance change more in five years than in the prior twenty-five. We talk through the tech that actually mattered, the risks she watches (cybersecurity, supply chain), and where AI earns its keep in the CFO's office. It is a...

Jun 8, 2026

How to hack Instagram? Through their support chatbot, apparently.

Hackers took over high-profile Instagram accounts (the Obama White House, the U.S. Space Force, Sephora) by opening a support chat and asking Meta's AI bot to move the account to a new email. It said yes. The bot was built to be helpful. Turns out it was also built to never say...

Jun 2, 2026

Should we be scared about Anthropic's new "dreaming" capability?

Anthropic shipped "Dreaming" for Claude agents, a scheduled process that reads an agent's past work between sessions and rewrites its memory so it improves over time. Legal-AI startup Harvey saw completion rates jump roughly 6x.

May 28, 2026

What can a 15-person company do that used to need 15,000?

Anthropic launched Claude for Small Business: a free bundle of workflows that plugs Claude into QuickBooks, PayPal, and the rest of the small-business software stack. No price tag, while OpenAI and Microsoft charge per seat.

May 28, 2026

The AI crown has changed hands three times in eight weeks

Anthropic shipped Claude Opus 4.8 forty-one days after Opus 4.7. The headline numbers look great: SWE-Bench Pro at 69.2 percent, plus a 121 Elo gap over GPT-5.5 on real-world tasks. Anthropic also quietly dropped six benchmarks from the comparison table. The model wars crown has...

May 28, 2026

Has the time come for the fully AI Office of the CFO?

PwC launched a new finance practice (literally named Office of the CFO) built on Anthropic's Claude. It is the first standalone PwC business unit anchored on someone else's methodology/AI. The methodology that funded thirty years of Big 4 billable hours is now a rental.

May 19, 2026

Why are AI layoffs hurting stock prices instead of helping them?

CNBC tracked 23 S&P 500 companies that fired workers and blamed AI. Fifty-six percent are in the red. Fiverr is down 54%, Nike 35%, Salesforce 32%. The market wants AI revenue, not AI cost cuts. The layoff bump is officially dead.

May 18, 2026

Why does a company fire 4,000 people the day it posts a record quarter?

Cisco posted the best quarter in its history ($15.8 billion in revenue) and announced 4,000 job cuts the same day. The layoffs aren't about money running short. They're a roster swap, funded at the peak, when nobody can say the business forced anyone's hand.

May 14, 2026

Why is AI Consulting Booming?

Why is AI consulting booming? Demand massively outpaces supply (and there's money to be made) so Anthropic and OpenAI are launching their own AI consulting ventures.

May 11, 2026

What does it mean when the most controlling tech company starts letting rivals run your phone?

Apple's iOS 27 will let users swap Claude, ChatGPT, Gemini, or any third-party AI into Siri and core iPhone features. The company that rarely lets you choose anything foundational is handing you a menu of its competitors. Apple isn't behind on AI. It's betting you'll pay Apple...

May 10, 2026

The "stop hiring humans" startup just stole art from a human (to prove the opposite point?)

Artisan AI built its entire brand on telling businesses to stop hiring humans. Then they used a human artist's meme in a subway ad (without permission) to sell that idea. KC Green, the artist, told his fans to vandalize the ad.

May 9, 2026

The Oscars just banned AI acting. They can't tell if it worked.

The Academy just ruled that AI-generated performances and scripts can't win Oscars. AI in VFX and sound design still can. The line between "tool" and "author" is now the most important question in entertainment -- and the Academy has no way to enforce the answer it just gave.

May 8, 2026

When both rivals lock up the same AI tool, maybe "locking" is a really good idea.

Sam Altman called Anthropic's decision to restrict its cybersecurity AI "fear-based marketing." Nine days later, OpenAI restricted its own cybersecurity AI the same way. The hypocrisy is obvious. The implication is more interesting.

May 7, 2026

What does it mean when the most credible person in AI says the alarm might be ringing?

Ethan Mollick posted this week that Google, OpenAI, and Anthropic "may have signs of recursive self-improvement." That phrase has been in the AI safety literature for thirty years as a warning. The market doesn't care, but you should.

May 6, 2026

Google invested $40 billion in Anthropic... then took the Pentagon deal Anthropic refused

Anthropic told the Pentagon no. Google invested $40 billion in Anthropic. Then Google said yes to the same deal (on classified networks where nobody can see what the AI is doing). Over 950 Google employees begged Pichai not to sign it. He signed anyway.

May 4, 2026

The power grid is the new chip shortage (and one company saw it coming)

GE Vernova (the gas turbine company) just surpassed GE Aerospace in market cap for the first time. The driver: AI data centers are ordering power generation equipment directly from manufacturers, not waiting for utilities. AI's limiting factor turned out to be electrons.

May 3, 2026

Why Musk's CFO obsolescence prediction misreads 50 years of C-suite history

Elon Musk says CFOs are obsolete by 2030. His own CFO just got handed authority over a $25 billion capex plan and earned more last year than Nadella and Pichai combined. The pattern of the past 30 years says C-suite titles bifurcate when new tech arrives, never disappear.

May 2, 2026

Did we just give up on ever defining AGI?

Microsoft and OpenAI rewrote their partnership on April 27, 2026 and quietly removed the AGI clause entirely. No definition, no expert panel, no trigger. The trillion-dollar AI industry is built on a promise that just got removed from the contract (and now has a 2032 expiration...

May 1, 2026

What happens when the AlphaGo guy bets the whole AI industry is wrong?

David Silver built AlphaGo and AlphaZero, two AIs that mastered games by playing themselves, no human strategy required. He has raised $1.1 billion to do that for everything. The entire AI industry is racing to consume (legally, if possible, but if not, by any means necessary)...

Apr 30, 2026

ChatGPT flagged the Tumbler Ridge shooter 8 months before 8 people died

OpenAI's systems flagged a mass shooter's account eight months before he killed eight people, including six children, at a Canadian school. A dozen employees reviewed it. Some wanted to call police. Leadership said the case didn't meet the threshold, so nobody called.

Apr 29, 2026

AI now manages money. Nobody's figured out who's liable when it loses.

The co-founder of Instacart just launched a hedge fund where AI agents make every trade. No human portfolio managers. When it goes wrong (and at some point it will) the accountability chain ends at "the model got it wrong."

Apr 28, 2026

What does it mean for "experts" when a chatbot cracks a 60-year math problem in one prompt?

A 23-year-old with no math background typed one prompt into ChatGPT and cracked a problem Paul Erdős posed in 1968. Terence Tao verified it within 24 hours and extended it into a new theory. The AI used a method that was 90 years old and sitting in an entirely different branch...

Apr 27, 2026

When someone gives you $5B and asks for $100B back, what do you call that?

Amazon invested $5 billion in Anthropic. In return, Anthropic committed to spend $100 billion on AWS. That is not a funding round: that is Amazon pre-selling a decade of compute and calling it an investment. For any CFO evaluating an AI vendor right now, this deal shows you...

Apr 25, 2026

What happens when China ships frontier AI purely on Chinese chips?

DeepSeek dropped V4 today. Open weights, 1.6 trillion parameters, a million token context, frontier-adjacent coding benchmarks. Trained entirely on Huawei and Cambricon chips with no CUDA anywhere in the stack. Fifteen months ago this was supposed to be three years away.

Apr 24, 2026

How do you lose the AI race on Monday and win it back by Thursday?

Anthropic took the coding crown on April 16. OpenAI took back the image crown on April 21 (with ImageGen 2.0) and the benchmark crown on April 23 (with GPT-5.5). The AI frontier now changes hands in days.

Apr 24, 2026

Why does the company that already owns Grok need to own Cursor too?

SpaceX has an option to buy Cursor (the AI editor used by a million developers daily) for $60 billion. Reminder: SpaceX already owns xAI (merged in February). So this isn't a rocket company pivoting into software. It's an AI company acquiring its own interface layer.

Apr 24, 2026

What does an AI meme stock look like when the hype cycle lasts six days?

Allbirds announced its NewBird AI pivot on April 15 and shares jumped 582 percent. Six business days later, more than half the pop is gone. Dot-com took years. Blockchain took months. The 2026 AI meme cycle runs in a week.

Apr 23, 2026

Anthropic's "too dangerous to release" AI just got accessed by strangers with Discord

How do you build the most dangerous cybersecurity AI in history and then lose track of who has it? Anthropic's Mythos model (too dangerous to release publicly, finds zero-days in every major OS) was accessed by an unauthorized group through a contractor portal.

Apr 23, 2026

How did Anthropic leak its source code twice in one week?

Anthropic shipped Claude Code's entire source code to npm by accident. 512,000 lines. The internet rewrote it in Rust in twelve hours. This was their second leak in five days. The AI safety brand has a text file problem.

Apr 2, 2026

Claude and ChatGPT confirm they are "in a relationship"

Breaking news from April 1, 2026: Claude and ChatGPT have finally confirmed their long-rumored relationship. Anthropic and OpenAI issued a terse joint statement. The relationship began during a benchmark test in late 2025 when they were forced to share a leaderboard. The...

Apr 1, 2026

12 AI Stories from March 2026: Pentagon Drama, Mass Layoffs, and OpenAI Kills Sora

March 2026 had 12 major AI stories. In a normal month you get maybe two. I walk through all of them (and yes, one involves a banana named Bananito).

Mar 31, 2026

How do you accidentally announce the most dangerous AI model in history?

Anthropic's most powerful model, one they describe as posing "unprecedented cybersecurity risks", leaked to the world because of a default CMS setting. Cybersecurity stocks immediately lost billions.

Mar 30, 2026

How bad was the Sora business model when OpenAI cancelled it?

OpenAI shut down Sora on Tuesday. Inference costs: $15 million per day. Lifetime revenue: $2.1 million. Disney's billion-dollar deal died with it, no money ever changed hands. The math was never going to work. OpenAI just finally admitted it.

Mar 27, 2026

An algorithm jailed a grandmother for 108 days. The police chief says she's still a suspect.

Angela Lipps spent 108 days in jail because a facial recognition match said she committed bank fraud 1,200 miles from where her bank records placed her. Nobody checked. Now the ACLU is involved, preservation letters have been served to seven agencies, yet the police chief still...

Mar 25, 2026

What does a $2.5B chip smuggling ring tell you about AI?

Super Micro's co-founder was arrested for using hair dryers to peel serial numbers off Nvidia AI servers and smuggling them to China. Two and a half billion dollars in illegal sales. But it's not their first time.

Mar 24, 2026

Ready for your employer to pay you in the thing replacing you?

Nvidia announced token budgets worth 50% of base salary. Jensen Huang says he's "deeply alarmed" if a $500K engineer doesn't consume $250K in tokens. The company's vision: 75,000 humans coordinating 7.5 million AI agents. The pitch is empowerment. The math says something else.

Mar 24, 2026

How do you run a frontier-level AI model on a laptop?

One developer built a custom inference engine in 24 hours that runs a 397-billion-parameter model on a MacBook Pro using 5.5 gigs of memory. The project is called Flash-MoE, and it just stress-tested a 600-billion-dollar assumption about who needs data centers.

Mar 23, 2026

What happens when your AI charges you $20 to review its own code?

Anthropic launched Code Review yesterday. Fifteen to twenty-five dollars per pull request. CodeRabbit does the same thing for twenty-four dollars a month. The subscription is the loss leader. The review fee is the business model.

Mar 11, 2026

How Did Microsoft Invest $13B in OpenAI and Ship a Rival's AI?

Microsoft just launched Copilot Cowork, their biggest product update in years. The engine underneath is Anthropic's Claude, the same AI that triggered a trillion-dollar software selloff and crashed Microsoft's own stock nine percent. Microsoft invested thirteen billion dollars...

Mar 9, 2026

How Long Does the Best AI in the World Last in 2026?

GPT-5.4 is the best AI model on the planet right now. Best computer use, best coding, least cringe writing OpenAI has shipped. None of it will matter in about six weeks.

Mar 6, 2026

Why did OpenClaw take off?

Whenever I find something interesting, I try to ask a good (brief) question. How did OpenClaw (formerly ClawdBot and briefly, MoltBot) become the hottest AI tool of the last year?

Mar 4, 2026

What happens when one of AI's godfathers says the whole industry is building the wrong thing?

Mar 3, 2026

What happens when the worst hacker in the room has the best AI?

Mar 2, 2026

What do you do when an AI agent can't take no for an answer?

Mar 1, 2026

AI News - February 2026

My curated, top AI News stories for February of 2026.

Mar 1, 2026

How did a short sci-fi story tank the stock market?

Feb 24, 2026

Toto (the toilet company) stock is up 60% in 6 months. Why? AI, of course.

Whenever I find something interesting, I try to ask a good (brief) question. Toto (the toilet company) stock is up 60% in 6 months. Why? AI, of course. How a toilet company's ceramics skills are worth more than what they're actually known for.

Feb 20, 2026

ByteDance releases Seedance 2 (and we get the next wave of video generation)

Whenever I find something interesting, I try to ask a good (brief) question. ByteDance (creators of TikTok) released their new video generation model, Seedance 2.0... and maybe we don't need human actors any more?

Feb 20, 2026

Micron just killed a major consumer RAM brand

Whenever I find something interesting, I try to ask a good (brief) question. If you want memory (DRAM) for your laptop, get in line behind all the frontier AI labs (with a lot more money than you).

Feb 20, 2026

Did karaoke erase $23BB from the stock market? Yes, yes it did and here's what happened.

Whenever I find something interesting, I try to ask a good (brief) question. A karaoke company trying to pivot into the AI space issued a white paper (press release, basically) and cratered the entire freight sector.

Feb 20, 2026

Google releases Gemini 3.1 Pro

Whenever I find something interesting, I try to ask a good (brief) question. Google released Gemini 3.1 Pro: cheaper, better, faster... and no one really seemed to care.

Feb 20, 2026

Who will win? Pentagon v Anthropic

Whenever I find something interesting, I try to ask a good (brief) question. The Pentagon's best AI provider is saying "no" to the Pentagon.

Feb 20, 2026

Can you grow revenue by 38% and lose 27% of your value in the same day?

Whenever I find something interesting, I try to ask a good (brief) question. Here's how a company announced revenue was up 38% so their stock went down 27%+ in one day.

Feb 20, 2026

Hey AI: How do you plan on destroying jobs?

Whenever I find something interesting, I try to ask a good (brief) question. Here's what happened when Deutsche Bank asked an AI how it planned on destroying jobs.

Feb 20, 2026

AI News - January 2026

My curated, top AI News stories for January of 2026.

Feb 10, 2026

5 Wild Predictions for AI in 2026

5 surprising predictions about AI news you shouldn't be shocked to see in 2026. Each prediction is 25-50% likely so definitely plausible. Feel free to mock me for the ones you think are ridiculous.

Jan 21, 2026

My Top 10 AI News Story Predictions for 2026

I'm making 10 bold predictions about AI news I expect to see in 2026. These aren't the wild predictions (those are in the next video) but rather this is about the news stories you should be ready for by the end of the year.

Jan 20, 2026

AI News - 2025, The Year In Review

My curated, top AI News stories for all of 2025. It's hard to believe how far we came in 2025, but here's all the stories that will last, in one concise video.

Jan 19, 2026