Archives
All the articles I've archived.
-
Meta Quietly Put AI Where Businesses Already Talk to Customers
Meta says its business AI tools reached 10 million conversations per week. The company may not win the model discourse, but it owns distribution.
-
Stripe Just Built the Wallet AI Agents Were Missing
Stripe upgraded Link so autonomous AI agents can request payments without seeing raw credentials. Agentic commerce needed trust before it needed hype.
-
Anthropic Is Becoming a Financial Instrument With a Model Attached
Google, investors, and Wall Street firms are circling Anthropic with tens of billions in capital. Claude is no longer just a model company story.
-
David Silver Raised $1.1 Billion to Teach AI Without Human Homework
DeepMind veteran David Silver raised $1.1B for Ineffable Intelligence, a London lab betting reinforcement learning can create AI beyond human data.
-
The AI Chip Story Just Moved From GPUs to CPUs
Meta signed a massive AWS Graviton deal for agentic AI workloads. GPUs still train the models, but CPUs may run more of the everyday agent work.
-
DeepSeek V4 Is the Boring Kind of Terrifying: Cheap, Huge, and Almost There
DeepSeek V4 previewed on April 24 with 1.6 trillion parameters, 1 million token context, open weights, and pricing that attacks frontier AI economics directly.
-
GPT-5.5 Is Not a Better Chatbot. It Is OpenAI Moving Onto the Computer.
OpenAI released GPT-5.5 on April 23 with stronger coding, tool use, research, and computer-control abilities. The chatbot era is becoming the work execution era.
-
Anthropic Gave You Extra Credits. Then Took Them Back the Moment You Blinked.
Opus 4.7 uses more tokens for the same work. Anthropic says they increased rate limits to compensate. Users report saved credits vanishing on subscription changes. This is becoming a pattern.
-
States Are Banning Therapy Chatbots. The Federal Government Wants to Stop Them.
Maine sent a therapy chatbot ban to the governor. Missouri is moving on a similar bill. Meanwhile, the White House wants Congress to preempt state AI laws. Both sides have a point. Neither has a solution. This is the defining regulatory tension of 2026.
-
Stanford's AI Index 2026: Agents Score Half as Well as PhD Experts. Everyone Is Adopting Them Anyway.
The 2026 Stanford AI Index Report reveals a paradox. AI agents perform only half as well as PhD experts on complex tasks. Spending is in the hundreds of billions. Adoption is faster than the PC or the internet. The gap between capability and capital has never been wider, and it might not matter.
-
Q1 2026: $300 Billion in VC Funding. 4 Companies Got $188 Billion of It.
Venture capital hit $300B in Q1 2026, up 150% year over year. 80% went to AI. The four largest rounds in VC history all happened in the same quarter. This is not diversification. This is winner-take-most at a speed we have never seen before.
-
Eli Lilly Built Pharma's Most Powerful AI Supercomputer. That Is the Entire Point.
LillyPod is 1,016 NVIDIA Blackwell Ultra GPUs on a DGX SuperPOD. Eli Lilly did not hire a consulting firm to write an AI strategy deck. They built a supercomputer. The gap between companies that treat AI as a feature and companies that treat it as infrastructure is about to become permanent.
-
$64 Billion in AI Data Centers Blocked by the People Who Live Next to Them
142 activist groups across 24 states have blocked or delayed $64 billion in data center projects. The AI infrastructure problem is not supply chain. It is consent.
-
Claude Opus 4.7 Is the Least Interesting Thing Anthropic Is Shipping This Week
Opus 4.7 shipped: 13% coding improvement, 21% fewer document errors, 3.75MP vision, new xhigh effort level. But the model is not the story. The platform play is.
-
China Processes 140 Trillion AI Tokens Per Day. The West Is Not Ready for What That Means.
Hong Kong AI IPOs up 490%. Zhipu stock up 570%. The Chinese government coined a word for 'token' and declared it an economic unit. While Silicon Valley debates safety, China built an economy around it.
-
74% of AI's Value Goes to 20% of Companies. The Other 80% Are Doing It Wrong.
PwC surveyed 1,217 executives across 25 industries. The leaders generate 7.2x more value from AI. The difference is not the tools. It is whether you redesigned the work or just bolted AI onto the old way.
-
Florida's AG Is Investigating ChatGPT Over a Mass Shooting. This Was Always Coming.
A shooter asked ChatGPT how to disengage his shotgun's safety three minutes before opening fire at FSU. Florida launched an investigation. The real question is what kind of AI policing comes next.
-
Meta Just Went Closed Source. Nobody Should Be Surprised.
After years of championing open-source AI with Llama, Meta released Muse Spark as a closed model. The open-source era was strategy, not philosophy. Now it is over.
-
OpenAI Launched a $100/Month Tier. Anthropic Is Already Beating Them While Spending Less.
OpenAI hit $25B ARR and launched a premium tier at $100/month. Meanwhile, Anthropic quietly passed them at $30B ARR while spending 4x less on training. The AI revenue race is no longer a one-horse show.
-
Anthropic Built an AI That Finds Zero-Days in Everything. Then They Did Something Smart.
Claude Mythos Preview can find zero-day vulnerabilities in every major OS and browser. Thousands of them. Autonomously. Including a 27-year-old OpenBSD flaw. Instead of releasing it to the public, Anthropic launched Project Glasswing.
-
OpenAI, Anthropic, and Google Just United Against China. That Should Tell You Something.
Three companies trying to destroy each other agreed to cooperate on one thing: stopping Chinese firms from stealing their models. Anthropic documented 16 million unauthorized exchanges. This is the AI cold war going hot.
-
Claude AI Just Drove a Rover on Mars. The 'AI Can't Do Real Work' Argument Is Over.
For 28 years, humans manually planned every meter a Mars rover drove. Then Claude wrote the commands and Perseverance drove 456 meters across Mars in two days. JPL engineers found only minor changes needed.
-
Google Just Made Its AI Coding Assistant Free. Cursor and Copilot Should Be Nervous.
Gemini Code Assist now offers 180K completions per month and 240 daily chat sessions at zero cost. No credit card. Google is doing what Google always does: making the competition's product free.
-
Yann LeCun Raised $1 Billion to Prove Every AI Company Is Wrong
AMI Labs just closed the largest seed round in European history. LeCun is betting against LLMs entirely, building JEPA world models instead. Either he is a visionary or he is about to burn a billion dollars in public.
-
I Pay $110 a Month for Claude. I Now Get One-Fifth of What I Used to.
Anthropic silently reduced Claude Code quotas, degraded model quality, capped rule files without warning, leaked their own source code, and then mass-reported 8,100 GitHub repos by accident. This is what paying for AI looks like in 2026.
-
Oracle Is Laying Off 30,000 People While Building AI Datacenters as Fast as It Can.
Oracle is cutting up to 30,000 employees across the US and India while aggressively investing in AI infrastructure. The largest enterprise tech layoff of 2026 so far, and the pattern should be familiar by now.
-
Mistral Just Borrowed $830 Million to Buy GPUs. Not Equity. Debt.
France's leading AI lab took on $830 million in debt to buy 13,800 NVIDIA chips and build a datacenter near Paris. When companies start borrowing money to buy GPUs, the compute crisis has entered a new phase.
-
Shopify Is Betting Everything on AI Storefronts. Walmart Already Proved It Doesn't Work.
Shopify launched 'Agentic Storefronts' inside ChatGPT, Gemini, and Copilot. The same week, Walmart revealed AI checkout converted 3x worse than their website. Two companies, same thesis, opposite conclusions. And both are missing the point.
-
Cloudflare Built a WordPress Replacement in Two Months With AI. It Is Called EmDash.
Cloudflare launched EmDash, an open-source CMS built on Astro that sandboxes plugins so they cannot destroy your site. Two engineers built it in two months with AI agents. WordPress's 24-year reign just got its first credible challenger.
-
The OpenAI Graveyard: Every Deal and Product That Died.
Sora, the Disney billion-dollar deal, the Windsurf acquisition, Stargate, the in-chat checkout, GPT-4o, the Operator standalone, the AMD partnership. OpenAI announces like a kingdom and delivers like a startup.
-
Jack Dorsey Cut 40% of His Workforce. His Reason: AI Can Do Their Jobs.
Block just fired 4,000 people. Jack Dorsey did not blame the economy or restructuring. He said AI tools make smaller teams possible. But here is what nobody is saying: it is not intelligence replacing people, it is automation replacing procedures.
-
Robotics Startups Raised $1.2 Billion in a Single Week. The Money Has Made Its Decision.
Five robotics companies raised over $1.2 billion in seven days. Mind Robotics, Rhoda AI, Sunday, Oxa, Fauna Robotics. The venture capital world just declared that physical AI is no longer a research bet. It is an investment thesis.
-
Google Just Made Every AI Model 6x Cheaper to Run. Memory Chip Stocks Crashed.
TurboQuant compresses LLM memory from 16 bits to 3 bits with zero accuracy loss. 6x less memory, 8x faster inference. And the stock market panicked, because apparently nobody learned from the DeepSeek episode.
-
Every Frontier AI Model Just Scored Below 1% on a Reasoning Test. Humans Score 100%.
ARC-AGI-3 is the first interactive reasoning benchmark for AI agents. Gemini scored 0.37%. GPT-5.4 scored 0.26%. Claude scored 0.25%. Humans solve every single one. The gap is not closing.
-
Four Rival Open Source Projects Merged Into One. The Result Is 88% Smaller.
Video.js, Plyr, Vidstack, and Media Chrome had 75,000 GitHub stars combined and powered tens of billions of video plays. Their creators put ego aside and rewrote everything from scratch together.
-
A Security Scanner Got Hacked. Then It Infected Everything It Scanned.
Trivy, one of the most trusted vulnerability scanners in DevOps, was compromised. The attack spawned CanisterWorm, a self-propagating npm worm that used blockchain as a command server. 141 packages infected. The irony is brutal.
-
Google Says Quantum Computers Will Break Encryption by 2029. Your Data Is Already Being Stolen.
Google warned that quantum computers capable of cracking today's encryption could arrive by 2029. The scariest part: attackers are already harvesting encrypted data now, planning to decrypt it later. This is called 'store now, decrypt later.'
-
Sam Altman Stopped Overseeing AI Safety. He Is Building Datacenters Instead.
The CEO of the most powerful AI lab just handed safety oversight to someone else so he can focus on fundraising and datacenter construction. He is kinda right, and that is the terrifying part.
-
NASA Just Let an AI Drive on Mars. After 28 Years of Humans Doing It.
Perseverance completed the first-ever AI-planned Mars drives using Anthropic's Claude to analyze orbital imagery. For 28 years, human operators planned every meter. That era just ended.
-
Unsung Hero: Damian Milton and the Theory That Changed Everything
In 2012, an autistic researcher named Damian Milton proposed that communication breakdowns between autistic and neurotypical people are mutual, not one-sided. The research community took a decade to catch up.
-
Unsung Hero: One Person Holds the JavaScript Ecosystem Together
Sindre Sorhus maintains 1,100+ npm packages with billions of downloads per month. If he stopped tomorrow, a significant portion of the npm registry would break. He funds himself through GitHub Sponsors.
-
Llama 4 Scout Has a 10 Million Token Context Window. Let That Sink In.
Meta released Llama 4 with a 10 million token context window. That is roughly 20 full novels. The implications for how we use AI are enormous and nobody is talking about the right ones.
-
An AI Agent Nuked 2.5 Years of Production Data. The Lesson Is Not What You Think.
An engineer trusted Claude Code to update his education platform. A misconfiguration wiped 2.5 years of student work. 100,000 students affected. The real story is about guardrails, not AI.
-
OpenClaw Went From 9K to 250K GitHub Stars in Four Months. Then Its Creator Left.
An open-source AI agent called OpenClaw surpassed React on GitHub. NVIDIA called it 'to agentic AI what GPT was to chatbots.' Then the creator joined OpenAI. What happens now?
-
Unsung Hero: Sophie Wilson Designed the Chip in Your Phone
In 1983, a woman in Cambridge designed an instruction set. 250 billion chips later, it powers every smartphone, every tablet, and every Apple Silicon Mac on Earth. Her name is Sophie Wilson.
-
Anthropic Said No to the Pentagon. Then OpenAI Said Yes.
Dario Amodei drew a hard line against military use of Claude. Days later, OpenAI signed a classified defense deal. ChatGPT uninstalls spiked 295%. This is the moment the AI industry split in two.
-
Meta Is Firing 15,000 Humans to Build Better Machines
Meta plans to cut 20% of its workforce to offset $135 billion in AI spending. The stock went up. The metaverse is dead. The machines won.
-
OpenAI Shut Down Sora. It Cost $15 Million a Day to Run and Made $2.1 Million Total.
OpenAI killed its AI video generator after burning $15 million per day in compute costs against $2.1 million in total lifetime revenue. But the real story is not about Sora. It is about the entire AI industry running on subsidized magic.
-
Walmart Tried AI Checkout Inside ChatGPT. It Converted 3x Worse Than Their Website.
Walmart put 200,000 products inside ChatGPT with instant checkout. The conversion rate was one-third of their regular website. The first real data point that agentic commerce is not ready.
-
GitHub Will Train AI on Your Code Starting April 24. You Have to Opt Out.
Starting April 24, GitHub will use your Copilot interactions to train AI models. Free, Pro, and Pro+ users are enrolled by default. You have to actively opt out. Most people will not.
-
OpenAI Raised $110 Billion. Read That Number Again.
Amazon put in $50 billion. NVIDIA put in $30 billion. SoftBank put in $30 billion. The largest private funding round in history values OpenAI at $840 billion. More than most countries. What could possibly justify this?
-
ElevenLabs Hit $11 Billion in Three Years. Two Founders. Voice AI.
From a 2-person startup to an $11 billion valuation in roughly three years. ElevenLabs raised $500 million from Sequoia to build the future of synthetic voice. The speed is almost absurd.
-
Unsung Hero: The Man Who Built the Invisible Machine That Compiles All Your Code
Chris Lattner's master's thesis became LLVM, the most widely used compiler backend on Earth. He then created Swift, MLIR, and is now building Mojo. Every time you compile code, his work is running underneath.
-
Unsung Hero: The Professor Whose Free Lectures Trained a Generation of Database Engineers
Andy Pavlo runs CMU's database research group. His free YouTube lectures have trained more database engineers than most universities. His research influenced DuckDB, CockroachDB, and every modern analytical database.
-
Perplexity Shipped an AI That Lives in Your House. On a Mac Mini. For $200 a Month.
Perplexity's 'Personal Computer' is an always-on AI agent running on a physical Mac mini in your home. It accesses your local files, coordinates other AI systems, and has a kill switch. This is a new product category.
-
Unsung Hero: The Professor Who Open-Sourced Legged Robots for Everyone
Boston Dynamics gets the viral videos. Professor Sangbae Kim at MIT built affordable legged robots and gave the designs away for free. The entire quadruped robot industry traces back to his lab.
-
Waymo Raised $16 Billion and Is Targeting 1 Million Rides a Week. This Is No Longer an Experiment.
Waymo launched its 6th-generation autonomous driver, raised the largest AV round ever at $16 billion, and is expanding to 11 cities plus Tokyo and London. The science project phase is over.
-
HP Made Customers Wait 15 Minutes on Purpose. Even When Agents Were Free.
HP deliberately forced callers in five European countries to wait 15 minutes to push them toward 'digital self-service.' The recording said 'we are experiencing longer wait times.' It was a lie.
-
Microsoft's January Patch Broke Windows. Then They Patched the Patch. Then They Patched That.
Microsoft's January 2026 update caused boot failures, crashes, and blue screens across billions of PCs. They needed two emergency patches in ten days. This is not a one-time event. This is a structural problem nobody can fix.
-
Boston Dynamics Put Atlas on a Factory Floor. China Did That Years Ago.
Atlas is sorting roof racks at a Hyundai factory in Georgia. The headlines call it groundbreaking. Meanwhile, Xiaomi runs fully automated factories with the lights off. The real robotics race is not what you think.
-
ChatGPT Health: The Tool You Want, the Trade You Don't See
OpenAI launched ChatGPT Health to translate your medical records into human language. The value is real. The question is what they get in exchange. And the answer is more than you think.
-
Unsung Hero: The Man Who Keeps Breaking Every Antivirus on Earth
Tavis Ormandy finds critical vulnerabilities in the software that is supposed to protect you. Windows kernel. Norton. Sophos. LastPass. GnuPG. He has made your computer safer, and you have never heard of him.
-
Unsung Hero: The Man Who Designed the Protocol That Runs the Internet
You have never heard of BGP. It is the only reason the internet works. Every packet you send is routed by a protocol co-designed by Yakov Rekhter. He quietly retired. Nobody noticed.
-
Jensen Huang Said 'The ChatGPT Moment for Physical AI Is Here.' Is He Right?
NVIDIA's CES 2026 keynote was the biggest in the company's history. Vera Rubin platform, Nemotron models, Cosmos world simulator, autonomous vehicles. Jensen says physical AI has arrived. The evidence is mixed.
-
One Year After DeepSeek R1: What Actually Changed?
A year ago, a Chinese lab nobody was watching released a model that crashed NVIDIA's stock and challenged every assumption about how much compute you actually need. I was watching them before the explosion. Here is what I think really happened.