Four stories. No fluff. Straight talk.
GOODBYE FAANG, HELLO MANGOS: THE SIX COMPANIES THAT JUST QUIETLY TOOK OVER THE ENTIRE ECONOMY
For about fifteen years, if you worked in tech, finance, or really any adjacent field where people wear blazers and talk about disruption at conferences, you knew the acronym. FAANG. Facebook, Amazon, Apple, Netflix, Google. The five companies that basically ran the economy, filled up the stock portfolios, and employed half of every Stanford CS graduate who ever existed.
Well, pour one out for FAANG because it just died a quiet death, and its replacement sounds like it belongs on a tropical fruit platter. Say hello to MANGOS. Meta. Anthropic. Nvidia. Google. OpenAI. SpaceX.
The new acronym started going viral on X after TechCrunch picked it up from a couple of developers who noticed the tech landscape had fundamentally shifted. And they are not wrong. Look at where the real weight is sitting right now. SpaceX just completed the largest IPO in history at $1.77 trillion. OpenAI filed for its own IPO at an $852 billion valuation. Anthropic is right behind them. Nvidia has been the hottest stock in the market for two straight years. Google is still Google, just now turbocharged by AI products people are actually using. And Meta is spending so much money on AI infrastructure it makes your head hurt.
What is interesting about MANGOS compared to FAANG is what got swapped out. Netflix is gone, because streaming is now basically a utility nobody thinks about. Amazon falls off this particular list, though its cloud business is still enormous. Apple drops out despite being worth three trillion dollars, which tells you something about how the AI era has reshuffled the rankings. You can be the most valuable company in history and still not be considered the future.
What MANGOS has that FAANG never did is that almost every company in the acronym is either building the AI infrastructure, running the most powerful AI models, or controlling the compute that all of it runs on. Nvidia sells the chips everyone else needs. SpaceX now owns the biggest privately held GPU cluster in the world and is renting it to Anthropic and Google at nearly a billion dollars a month each. Google, OpenAI, and Anthropic are in a three-way fight for the model crown. And Meta is spending somewhere north of sixty billion dollars this year alone on AI because Mark Zuckerberg has decided open source AI is how he beats everyone else.
The darker implication nobody is saying out loud is that six companies are about to control most of what artificial intelligence produces and decides at a civilizational scale. FAANG was dominant. MANGOS is something different. These are not just big companies with a lot of users. These are the companies building the operating system for the next era of human civilization. Whether you find that exciting or terrifying probably says a lot about where you sit on the optimism spectrum.
One more thing worth saying. MANGOS sounds ridiculous and that is fine. Most of the most powerful things in history had names that sounded ridiculous when you first heard them. You get used to it.
Read the full story at TechCrunch.
THE APP THAT LET NON-CODERS BUILD SOFTWARE JUST HIT $500 MILLION AND NOBODY OUTSIDE TECH IS PAYING ATTENTION
Okay, so you are going to want to hear this one because it is quietly one of the most important business stories of 2026 and most people are not paying attention to it.
There is a company called Lovable. It is Swedish. It is barely three years old. And it just announced that it has crossed $500 million in annualized revenue while its users are collectively building one million new software projects every single week.
Let that number sink in. One million projects a week. Not users. Projects. People are using this thing to build actual software: e-commerce stores, CRM tools, inventory systems, HR platforms, websites, internal business apps. The majority of users are non-technical, meaning they have never written a line of code in their lives and they are now shipping functioning software products because Lovable lets you describe what you want and it builds it.
This is what the tech industry has been calling vibe coding, which is a phrase I genuinely dislike but which perfectly describes the experience. You describe your idea to an AI, it builds the thing, you tell it what is wrong, it fixes the thing. No computer science degree required. No Stack Overflow. No crying into coffee at midnight wondering why the variable names stopped making sense.
Lovable crossed $400 million in February. Now they are at $500 million. Back in August 2024, the founders said they thought they could hit one billion in annualized revenue within twelve months. They probably will not quite get there by summer, but the trajectory is still jaw-dropping for a company that has never had its third birthday.
Here is what matters about this story beyond the headline number. TechCrunch made a sharp observation in their coverage: the hard part of software is not building it. Building it has always been the fun part. The hard part is maintaining it. Software is a living organism. Dependencies update. APIs break. Third-party services change their pricing and suddenly your app stops working in the middle of the night and nobody knows why. That is why companies still pay Salesforce and Workday hundreds of millions of dollars a year, because they want someone else responsible for keeping the lights on.
Lovable has not existed long enough to tell us what happens when a million vibe-coded projects hit their first maintenance crisis. Do people fix them, or do they just build a new one? If the answer is mostly just build a new one, then Lovable has not built a software development platform. It has built a disposable software factory. Which might be fine, honestly, or it might create technical debt at a scale nobody has ever seen before.
Either way, a two-year-old Swedish startup is doing half a billion dollars a year in revenue by teaching the world that anyone can now be a software developer. That is not nothing. That is, in fact, a pretty big deal for every enterprise software company currently charging you for a subscription.
Read the full story at TechCrunch.
WARNER MUSIC JUST BOUGHT THE COMPANY THAT KNOWS EXACTLY WHEN AI STEALS YOUR SONG
If you have been following the war between the music industry and artificial intelligence over the last few years, you know it has been a mess. Lawsuits, settlements, angry open letters, more lawsuits, record labels suing AI startups, AI startups settling, everyone fighting over who owns what when a machine trains on your music and then spits out something that sounds just like you.
Well, Warner Music Group just made its most interesting move in the AI era. They acquired a startup called Sureel AI, and unlike most of the AI news you have read this week, this one is genuinely clever.
Here is the thing about Sureel. It does not just look at whether an AI-generated song sounds like an artist. That is a blunt instrument and a legal nightmare to prove in court. What Sureel has built is something they call AI DNA for songs. The technology breaks a piece of music down into its component parts, all the way down to the specific elements that make a song distinctly itself, and then tracks where those elements show up in AI-generated content or in the training data that went into an AI model. They also have a suite for tracking name, image, and likeness, meaning if an AI generates a voice clone that sounds like a specific artist, Sureel can identify where that clone originated and document the full chain.
This is a significant weapon for a major label to own. Warner was originally on the side of record labels suing Suno, the AI music startup, back in 2024. They settled with Suno last year and signed a licensing deal. They also settled with Udio, another AI music startup. So Warner is now in the business of making deals with AI music companies, and what they just bought is the technology to tell them exactly how much they are owed every time those deals get exercised.
Think of it like this. Music royalties work because someone has to account for every time a song plays, somewhere, and someone has to pay the rights holder. That whole system runs on tracking. Sureel is building that same infrastructure but for the AI age, where instead of a song playing on a radio station, your song might be embedded inside the training data of a model that generates a billion outputs a year.
Sony and Universal are still fighting Suno in court, doing things the old-fashioned legal way. Warner went ahead and bought the company with the technology to actually prove those cases. That is a different kind of bet, and honestly a smarter one. Instead of arguing in front of a judge about whether a song sounds like another song, Warner will be able to show exactly which frames of audio, down to the musical DNA level, ended up where.
The deal terms were not disclosed. Given that Warner is navigating a streaming market that is slowing down and an AI industry that is actively eating their lunch, the guess here is they got Sureel for a reasonable price and got a genuinely useful piece of infrastructure. The future of music industry revenue in the AI age will be tracking and licensing, not lawsuits. Warner just equipped itself for that future.
Read the full story at TechCrunch.
THIS STARTUP JUST BUILT A MACHINE THAT SIMULATES THE ENTIRE WORLD IN REAL TIME, AND SELF-DRIVING CARS ARE JUST THE BEGINNING
Here is a piece of technology news that sounds like science fiction but is very much real and worth your attention.
A startup called Decart just launched something called Oasis 3. It is what AI researchers call a world model, which is an ambitious name but basically accurate. What it does is generate photorealistic driving environments in real time, entirely from a text prompt. You type in something like “New York City street at 8 in the morning” and it builds a complete, visually convincing three-dimensional driving environment you can then navigate through. Not a video clip. Not a rendering that takes hours. A real-time, interactive world generating itself as you move through it.
The target customer right now is autonomous vehicle companies. Self-driving technology has always had a fundamental testing problem. You can only drive so many real-world miles before the cost becomes enormous and the time required becomes years. The edge cases, the truly weird stuff that causes accidents, happen rarely on real roads. You might drive a million miles without seeing the particular combination of weather, road condition, and pedestrian behavior that would seriously challenge your system. Oasis 3 lets you simulate that edge case infinite times in infinite variations until your model has seen it enough times to know what to do.
Decart raised $300 million in a recent round at nearly $4 billion valuation, with Toyota, Adobe, Nvidia, and eBay among the strategic investors. The model is available via API right now at $0.02 per second. The company says it is an order of magnitude cheaper to run than competitors because they built their software all the way down to the hardware level rather than stacking things on top of commodity cloud services.
The honest caveat, because TechCrunch did hands-on testing, is that the current version has real limitations. The photorealism holds up beautifully for the first few seconds. Then things start to drift. You turn around to go back to where you started and the intersection is gone, replaced by a different environment entirely. Objects do not have proper physics, meaning your simulated car will just drive through other simulated cars like a ghost. The model does not yet understand that two solid things cannot occupy the same space at the same time, which is, generally speaking, a fairly important property for a self-driving testing system to get right.
The CEO acknowledges all of this openly and says they are working on it. The memory problem specifically, where the model burns through context at hundreds of thousands of tokens per second, is a known research challenge. His comparison was to the early days of large language models, before the context windows grew from 2,000 tokens to a million tokens. The capability is there. The infrastructure to sustain it for longer is still being built.
But here is what is exciting despite the limitations. Decart is opening Oasis 3 to developers via API and saying, go build things on top of this. The founder compared it to what OpenAI did when they released language model APIs in 2020 and an entire ecosystem emerged that nobody at OpenAI had predicted. What happens when developers have access to a real-time world-generating machine? Applications in robotics, gaming, architecture, emergency response training, military simulation, urban planning are all plausible. The self-driving use case is just the first obvious door. There are a hundred more that nobody has tried to open yet.
Read the full story at TechCrunch.