{
 "creator": "Natalie Fratto",
 "handle": "@nataliefratto",
 "platform": "Instagram",
 "profile_url": "https://www.instagram.com/nataliefratto/",
 "method": "Top 10 of 49 Instagram reels on @nataliefratto, all time, ranked by Instagram play count. Sponsored posts are kept and marked (one reel carries a #databrickspartner tag). Her reels run 2 to 3 minutes, past ScrapeCreators' 2-minute transcript limit, so transcripts were made with Whisper large-v3-turbo from the reel audio and lightly corrected for misheard names.",
 "snapshot_date": "2026-09-24",
 "videos_considered": 49,
 "videos": [
  {
   "rank": 1,
   "title": "Databricks is the hardest company I've tried to explain yet.",
   "shortcode": "DZFnttWijxS",
   "url": "https://www.instagram.com/reel/DZFnttWijxS/",
   "creator": "Natalie Fratto",
   "handle": "@nataliefratto",
   "views": 6315391,
   "likes": 230987,
   "comments": 740,
   "published": "2026-06-02",
   "duration": "2:09",
   "sponsored": false,
   "transcript": "Databricks is the hardest company I've tried to explain yet.\nThis is part seven in my series explaining top private tech companies.\nFor decades, companies stored their data in one of two places.\nA data warehouse, which is perfect for structured data like spreadsheets and order forms.\nIt's fast and reliable, but expensive and rigid.\nOr a data lake, built for messier data like videos and images.\nIt's cheap and flexible, but slow and unreliable.\nMost large companies use both.\nTwo separate systems, two teams managing the whole mess.\nIn 2013, seven researchers from Berkeley built a way to process massive data sets by distributing the work across thousands of computers at the same time.\nThey called it Apache Spark.\nOn top of Spark, they built something called the Lakehouse.\nA hybrid of the two that combined the best of both.\nFast, reliable, cheap and flexible.\nOnly possible because it was built on Spark.\nThen they made a strange decision.\nThey open sourced both Apache Spark and the Lakehouse, giving it away entirely for free.\nThey bet that if the world built on top of their technology, the adoption would create a moat.\nIt totally worked.\nOver the next decade, millions of developers across thousands of big companies built their entire data infrastructure on Spark.\nDatabricks instead sold proprietary features and managed platform service on top of this open source foundation.\nIn 2022, when ChatGPT was released, Databricks was pretty perfectly positioned.\nTheir customers could now build AI directly on top of the data they already had, using Databricks for retrieval, model training, and autonomous agents.\nAnd then pay Databricks like a utility bill.\nThe more AI workloads they ran, the higher the utility bill.\nDatabricks' existing infrastructure just became infinitely more valuable because of AI.\nThe revenue growth is wild, from $200 million in 2020 to a $5.4 billion run rate in 2026, roughly 27x in six years, with the curve steepening every year after ChatGPT.\nOver that period, Databricks raised $20 billion, from Andreessen Horowitz, Thrive, NVIDIA, and others, with a current valuation of $134 billion.\nAnd IPO is on the horizon.\nIf data really is the new oil, Databricks owned the pipeline, and they built it by letting the oil flow freely.",
   "caption": "Databricks is on every list of hot companies, but before making this video, I\u2019m not sure I could have explained it to you with a gun to my head. Did the research for all of us - here\u2019s the 2 min pretty version!\n\n#databricks #developer #technology #ai #startup",
   "page": "https://creator-examples.vercel.app/short-form/natalie-fratto/01-databricks-is-the-hardest-company-i-ve-tried-to-explain-yet/"
  },
  {
   "rank": 2,
   "title": "This is the weirdest company winning in AI, part two.",
   "shortcode": "Dbq1_X8Jg7Z",
   "url": "https://www.instagram.com/reel/Dbq1_X8Jg7Z/",
   "creator": "Natalie Fratto",
   "handle": "@nataliefratto",
   "views": 4078580,
   "likes": 225258,
   "comments": 1310,
   "published": "2026-08-05",
   "duration": "2:55",
   "sponsored": false,
   "transcript": "This is the weirdest company winning in AI, part two.\nLast time we talked about toilets, and this time we're talking about the people who make MSG.\nAjinomoto, the Japanese seasoning company that makes your food taste savory, also makes a tiny film layer that's almost invisible, and yet basically every high-performance chip wouldn't work without it.\nLet me explain.\nIn order to get power into chips and data out of them, you need to connect them to PCBs.\nIn the olden days, you'd add metal legs around the edges of chips, but as chips advanced, you needed more and more, to the tune of thousands, so we ran out of space.\nInstead, they started using multiple layers of copper wiring stacked underneath the chip, between it and the board.\nThis is called packaging substrate.\nIn order to have the copper not hit each other, you needed to insulate it with something.\nThat insulation layer has a bunch of really strict requirements.\nIt needs to be drillable to withstand thousands of perfect microscopic holes.\nIt can't move or expand too much when it gets heated up.\nIt has to be invisible to electricity.\nAnd it needs to stay flawless at big lengths.\nThe film is thinner than one human hair and gets shipped in huge rolls that need to stay intact.\nFor years, the industry used ink between these layers of chips, but it went on unevenly and trapped air bubbles.\nMeanwhile, this Japanese food ingredient company, Ajinomoto, had been selling MSG since 1909.\nAround 1970, MSG had a big PR challenge, as people linked it to health scares.\nSo the company looked around what they had for R&D expansion opportunities beyond this single product.\nLucky coincidence, MSG's chemical structure is eerily similar to the plastics used inside electronics.\nThey're basically cousins.\nSo they studied it for 20 years with nothing to sell.\nThen finally, in 1996, a CPU maker knocked on their door, basically saying, Hey, this ink is failing us.\nCould you make us something better?\nAjinomoto did a four-month sprint and came up with this insulating film that was just perfect.\nFor the next 25 years, they held a 95% monopoly across the space.\nAnd their insulating film is found inside virtually every single personal computer and chips made by Intel, NVIDIA, AMD, and more.\nIt's a sneaky good business, too, because it's super hard to copy.\nThey hold hundreds of patents and they co-develop each film layer alongside the chip maker.\nSo they get insider intel on all the requirements up to five years earlier than any competitor.\nOver the last few years, the AI boom has meant bigger chips, more layers, and more film.\nAjinomoto's chip materials profit jumped 35% last year to about $350 million, now 30% of the entire company's profit, at margins above 50%.\nThey just raised prices, 30%, and nobody blinked an eye.\nIt's kind of the same story as toilets.\nAn old, boring product built the empire, but now a previously invisible side quest is printing all of the money.",
   "caption": "Not that kind of chips.\n\nFirst Charts & Crafts video with a prop \ud83d\ude0f this one was so fun to learn about.\n\n#ai #semiconductors #technology #tech #explainer",
   "page": "https://creator-examples.vercel.app/short-form/natalie-fratto/02-this-is-the-weirdest-company-winning-in-ai-part-two/"
  },
  {
   "rank": 3,
   "title": "I'm going to try to explain the entire AI stack in just two minutes.",
   "shortcode": "DZ7lsSBRj5Q",
   "url": "https://www.instagram.com/reel/DZ7lsSBRj5Q/",
   "creator": "Natalie Fratto",
   "handle": "@nataliefratto",
   "views": 1696862,
   "likes": 49595,
   "comments": 261,
   "published": "2026-06-23",
   "duration": "2:41",
   "sponsored": true,
   "transcript": "I'm going to try to explain the entire AI stack in just two minutes.\nLock in with me because an entirely new layer just got added this week.\nIt helps me to think about AI like a new city being built.\nAt the very base, we have the models, the big ones you know.\nOpenAI's ChatGPT, Anthropic's Claude, Google's Gemini, and more.\nThese are the raw plots of land, the technical soil everything else grows from.\nIf you're a developer, you can build on top of this land directly, but for most people, something has to be built on top of it before it becomes useful.\nOn top of those models are harnesses.\nThink of these like storefronts, often made by a model company as a way to make their model actually do stuff.\nWhile models can only talk, harnesses take actions.\nThey read files, write code, run commands, search the web, remember prior history.\nMost people interact with AI through these harness storefronts.\nAs a user, you can visit any harness to get your work done, but ultimately they're disconnected.\nNone of your workflows, context, or special configurations travel between the different plots of land.\nIf you use Claude Code today and have a better tool that launches tomorrow, you have to start all over on that new land.\nFor enterprise teams, it's even worse.\nTwo people on completely separate plots of land with no way to reach each other, no idea what the other person's agent is doing, no idea what it's costing.\nBut Databricks just released something big called Omnigent to fix that.\nThe way I think about it is they finally built a highway system to connect everything in AI.\nThey're calling Omnigent a meta-harness.\nYour work, your history, your settings can now travel with you regardless of which harness you're using.\nA previously isolated plot of land now becomes a full city grid.\nFor enterprise teams, this means you can share your AI session with a teammate just by sending a link, kind of like you would share a Google Doc.\nYou can also track your spend together and set security policies across your whole team.\nNow that we have this new city grid, people are going to start building more on top of it.\nDatabricks itself built two examples.\nPolly, which breaks a task into pieces and sends them to different agents to work in parallel, then has each piece checked by a different model than the one that wrote it.\nAnd Debbie, which sends every question to both Claude and GPT at the same time, so you can compare the answers.\nBut anyone else can build on it too, even Databricks competitors.\nThat part is key.\nFour million of you watched our last Databricks video, where we learned that the critical business decision they made was giving away Spark for free.\nWell, they're doing it again.\nOmnigent is fully open source, and the bet is by doing so, everyone, everywhere, will use it.\nAs this whole AI city grows, they're positioning Omnigent as critical infrastructure.\nAnd history tells us, in any new city, it's pretty smart play to be the road builders.",
   "caption": "Agent this, harness that. Here\u2019s the entire AI stack broken down, plus a whole new open source layer introduced by @databricksinc last week called Omnigent that changes the game entirely \n\n#databricks #ai #omnigent #databrickspartner",
   "page": "https://creator-examples.vercel.app/short-form/natalie-fratto/03-i-m-going-to-try-to-explain-the-entire-ai-stack-in-just-two-minutes/"
  },
  {
   "rank": 4,
   "title": "Silicon Valley has changed everything, even the way we go to war.",
   "shortcode": "DX9bBZKCpVm",
   "url": "https://www.instagram.com/reel/DX9bBZKCpVm/",
   "creator": "Natalie Fratto",
   "handle": "@nataliefratto",
   "views": 1141979,
   "likes": 54496,
   "comments": 555,
   "published": "2026-05-05",
   "duration": "2:20",
   "sponsored": false,
   "transcript": "Silicon Valley has changed everything, even the way we go to war.\nThis is part four in our series explaining the tech and business behind the most valuable private tech companies.\nThe Anduril founder, Palmer Luckey, is storied.\nAt just 20, he sold his VR headset company, Oculus, to Facebook for $2 billion.\nTwo years later, he was fired by Zuckerberg and partnered up with Trey Stevens from Founders Fund to build a defense startup they likened to Stark Industries from Iron Man.\nThe name Anduril comes from Aragorn's sword in Lord of the Rings and means Flame of the West.\nThey aimed to turn the defense business on its head.\nTraditionally, it's built to spec, and it works like this.\nThe government puts out a proposal, big contractors, Lockheed, Boeing, Raytheon, bid on it, the winner gets paid up front, and then the product is built.\nInstead, Anduril flipped it.\nThey call their approach built to mission, where they take on R&D risk, build working systems using their own VC dollars, and then sell finished products off the shelf to the government.\nTheir technology is software-defined, hardware-enabled.\nAt the core, there's Lattice, a software product that's like a brain, and all other hardware products run on it.\nThose include Sentry Tower, their first product, a foldable, portable, solar-powered surveillance tower.\nUAVs like Ghost Drone, a near-silent autonomous helicopter, and Roadrunner, described as somewhere between a drone and a missile, designed to take down incoming threats and land itself.\nFury, a new unmanned fighter jet that will soon fly alongside the U.S. Air Force, and IVAS, an AR headset for the U.S. Army.\nThis one was kind of full circle for Luckey, because it's close to his VR roots, and it was a contract Anduril recently won away from Microsoft.\nAnduril has raised $6.26 billion from VCs including Founders Fund, Andreessen Horowitz, General Catalyst, and Thrive.\nThe company is valued at $30.5 billion, 14 times revenue.\nNot profitable until 2030.\nAnduril's business is naturally divisive, given the nature of their products.\nThere are kind of two camps, Anduril and their investors, who believe that avoiding defense products is both a moral failure and a missed market opportunity, and those who think letting for-profit companies set the defense product suite sets a risky precedent.\nEither way, in only seven years, Anduril has built the most important defense contractor in America by building weapons like Silicon Valley software.\nWhich company on this list do you want to see unpacked next?",
   "caption": "Your most highly requested Charts & Crafts topic: @anduril \n\n#anduril #palmerluckey #defensetech #nationalsecurity #tech",
   "page": "https://creator-examples.vercel.app/short-form/natalie-fratto/04-silicon-valley-has-changed-everything-even-the-way-we-go-to-war/"
  },
  {
   "rank": 5,
   "title": "Energy is getting really scarce for AI, but there's one company that turned that problem into a...",
   "shortcode": "Dcd0nZkCnAa",
   "url": "https://www.instagram.com/reel/Dcd0nZkCnAa/",
   "creator": "Natalie Fratto",
   "handle": "@nataliefratto",
   "views": 755246,
   "likes": 20106,
   "comments": 123,
   "published": "2026-08-25",
   "duration": "2:21",
   "sponsored": false,
   "transcript": "Energy is getting really scarce for AI, but there's one company that turned that problem into a super surprising advantage.\nCrusoe is just eight years old, and it's now one of the most valuable private tech companies.\nHere's the crux of the energy problem.\nOur demand for AI is running way ahead of what our entire energy grid can deliver.\nThe densest cluster of AI data centers on Earth is in northern Virginia, about four gigawatts.\nAnd AWS, Azure, and everyone else in the hyperscaler category is plugging in there, but it's already full.\nIf you wanted to add another 100 megawatts, the average wait time is seven whole years.\nCrusoe proposed a better way, moving you off the grid waitlist entirely by finding hidden energy sources that nobody else can use.\nThe company was started in 2018 by two friends, Chase from Quant Trading and Cully from the energy sector.\nTheir first product used trapped energy from oil fields, where methane gas was being burned off into the air.\nThey rigged up shipping container-sized boxes with generators and computers to sit nearby, capture that methane gas, and use it to power Bitcoin mining.\nPretty solid business.\nTurns out, AI training also has similar power needs.\nSo in 2024, Oracle put out an RFP for 100 megawatts of power for an unnamed customer.\nThis turned out to be OpenAI.\nPeople lined up to bid and claimed they could cut seven years down to two.\nChase said he was the 35th person to get that call, nowhere near the top, but he said he could get them that power in just one year.\nBy building it in Abilene, Texas, using a similar idea.\nWest Texas is wind country, and decades of tax credits built up a lot of wind turbines out there.\nBut that same grid clog meant that about a fifth of Texas renewable power never made it into supply.\nIt's trapped.\nSo rather than move the power to the demand, Crusoe moved the demand to the power.\nThey won the bid and built a huge campus.\nIt's 90 acres, 4 million square feet, and they've had to add about a gigawatt of power from gas turbines, too, just to meet the speed requirements needed.\nWhen complete, it'll draw about a third as much energy as the whole northern Virginia region in just one campus.\nCrusoe raised last at a $10 billion valuation in October, and reportedly they're in talks to raise at $30 billion right now, with rumors of an IPO on the horizon.\nWe'll see if this trapped energy play can take them all the way.",
   "caption": "Top private tech companies, part 9: Crusoe \n\n#crusoe #energy #datacenter #technology #ai @crusoe.ai",
   "page": "https://creator-examples.vercel.app/short-form/natalie-fratto/05-energy-is-getting-really-scarce-for-ai-but-there-s-one-company-that-tu/"
  },
  {
   "rank": 6,
   "title": "This one's a little weird. There's a toilet manufacturer in Japan that's sneakily critical to the...",
   "shortcode": "DbGTYDQxg4f",
   "url": "https://www.instagram.com/reel/DbGTYDQxg4f/",
   "creator": "Natalie Fratto",
   "handle": "@nataliefratto",
   "views": 751200,
   "likes": 29632,
   "comments": 242,
   "published": "2026-07-22",
   "duration": "2:41",
   "sponsored": false,
   "transcript": "This one's a little weird.\nThere's a toilet manufacturer in Japan that's sneakily critical to the entire chips manufacturing process, and they're kind of flush with demand right now.\nI was in Japan last week, and a friend told me a really wild story.\nSo it all starts back in the early 1900s with this Japanese founder who went on a trip to Europe and was really impressed by their ceramic toilets.\nHe returned back to Japan to start a company building toilets to replace the wooden ones Japan had up until then.\nCeramic is a surprisingly complex material, especially for a toilet.\nToilets need to be really, really smooth for sanitary reasons, have a low error tolerance so all the piping and plumbing fits and it won't leak, and then you have to shape the whole thing first, bake it in a kiln where it shrinks 13% to exactly the right dimensions.\nIt took them over 17,000 tests to get the first good toilet.\nSo the company grows around the country.\nThen in the 1980s, Japan leads the world in the first semiconductor boom.\nPlayers like Toshiba and Hitachi emerge.\nAt its peak, Japan is responsible for 50% of global chip production.\nToto wants to find a way in, so they established their advanced ceramics division in 1984.\nSoon after, they partner with NTT Laboratories, which is Japan's Bell Labs equivalent, to make a new product called the Chuck.\nHere's what that is.\nWhen you make a semiconductor, you blast plasma inside a vacuum chamber down onto a silicon wafer in order to etch it with the designs needed for computations.\nThese etchings are incredibly precise, and it's important that the wafer stays in place, doesn't bend, and doesn't snap.\nSo underneath it, you want to put a chuck to hold it still.\nChucks can't use metal or have prongs because it would leave residue shavings behind, but it has to be really sturdy.\nSo ceramic is actually perfect because it's incredibly strong, but it also can be connected via electrostatic forces, the same static energy that sticks a balloon to your hair.\nSo Toto builds it, releases it, and it kind of flops for 30 years until the AI boom over the last few years.\nAs microchips become more complex and memory-focused, we've seen a totally new kind of etching and chip take off, called 3D NAND.\nThese are basically much deeper chips that stack memory cells vertically instead of horizontally.\nTo make them work, you etch holes straight down through every layer of the chip 50 times deeper than they are wide.\nThis process makes it even more critical that the wafer stays in place and intact, so chucks become even more important.\nToto is the world's number two chuck maker and the key supplier for the machines that etch 3D NAND chips.\nAs a result, their advanced ceramics division is finally booming.\nThis year, it accounted for 57% of their total operating profit, way more than the toilets.\nWhat's the opposite of a business going down the drain?",
   "caption": "Weirdest winner in semiconductors award goes to: TOTO electronics, maker of the ceramic \u201cchucks\u201d used in the most advanced memory chip machines \n\n#chip #semiconductor #technology #ai #electronics",
   "page": "https://creator-examples.vercel.app/short-form/natalie-fratto/06-this-one-s-a-little-weird-there-s-a-toilet-manufacturer-in-japan-that/"
  },
  {
   "rank": 7,
   "title": "The SpaceX origin story? It's even crazier than you know.",
   "shortcode": "DXnUSFTDd7w",
   "url": "https://www.instagram.com/reel/DXnUSFTDd7w/",
   "creator": "Natalie Fratto",
   "handle": "@nataliefratto",
   "views": 500012,
   "likes": 27053,
   "comments": 389,
   "published": "2026-04-27",
   "duration": "2:04",
   "sponsored": false,
   "transcript": "The SpaceX origin story?\nIt's even crazier than you know.\nThis is part three in our Charts & Crafts series on top private tech companies.\nIn 2001, Elon Musk had an idea to launch a greenhouse to Mars, so he went to Moscow trying to buy a refurbished ICBM from the Russians.\nThey said no, and on the flight home, he decided to make the rocket himself, and make it reusable.\nHere's how a reusable rocket works.\nOnce it launches, the booster, bottom part, separates from the upper stage, and while the payload, stuff on it, continues on its course, the booster flips itself around using nitrogen thrusters, going further up and then back down again, burning the engines two more times to very slowly re-enter and touchdown vertically on a drone-operated ship in the ocean.\nIt's all autonomous.\nThe rocket lands itself.\nIt took 14 years to perfect this technology, but now each booster has 20 or so uses before repairs.\nReusing rockets is way cheaper.\nNASA's SLS costs about $1.5 billion per launch.\nSpaceX's Falcon 9 is $67 million, 20 times less.\nWhen you look at total 2025 launches globally, their dominance is clear.\nSpaceX launches account for the vast majority of the total, followed by China and then everyone else.\nOkay, so you'd think these launch services are a majority of the revenue, right?\nWrong.\nThe vast majority actually comes from what's on the payload, Starlink satellites.\nTheir sneaky huge telecom business.\nMore than two-thirds of last year's revenue came from Starlink, and only a third came from launch services.\nStarlink's business is deeply intertwined with the launches, though.\nTogether, they're a flywheel.\nCheap, reusable rockets means more satellites, which means R&D from higher revenue, which means even more cheap, reusable rockets.\nSpaceX currently sits at a trillion-dollar valuation and has only raised $12 billion to date.\nTheir valuation is 90 times their revenue.\nThey filed for an IPO this month, and if successful, it's going to be the largest IPO in history, by more than double.\nWhich company on this list do you want to see unpacked next?",
   "caption": "It\u2019s actually a telecom business wrapped in a rocket company\n\n#spacex #elonmusk #tech #space #vc",
   "page": "https://creator-examples.vercel.app/short-form/natalie-fratto/07-the-spacex-origin-story-it-s-even-crazier-than-you-know/"
  },
  {
   "rank": 8,
   "title": "You're so wrong if you think Stripe is just a payments company.",
   "shortcode": "Dba3INFi-0v",
   "url": "https://www.instagram.com/reel/Dba3INFi-0v/",
   "creator": "Natalie Fratto",
   "handle": "@nataliefratto",
   "views": 428895,
   "likes": 14010,
   "comments": 102,
   "published": "2026-07-30",
   "duration": "2:48",
   "sponsored": false,
   "transcript": "You're so wrong if you think Stripe is just a payments company.\nIt's a whole self-contained universe.\nThis is part eight in our Charts & Crafts series explaining top private tech companies.\nI never quite understood how literal Stripe's motto, increasing the GDP of the internet, was.\nLet me explain.\nStripe was founded by two brothers who grew up in rural Ireland, Patrick and John Collison, to fix a pain point from their earlier company, accepting payments online.\nBefore Stripe, if you wanted to sell t-shirts online, you'd build a website, then grind through a set of five or more clunky steps behind the scenes.\nYou'd set up a merchant account, which takes weeks of paperwork, install a payment gateway, plug into a payment processor, pass PCI compliance, and add a fraud detection system.\nTheir solution was to put all of that mess into a tidy single product, one any developer could install with just seven lines of code.\nThis created a boom in economic activity because suddenly, far more people could sell things online.\nThe scale of that boom is kind of mind-blowing.\nIn 2024, 1.4 trillion flowed through Stripe, about 1.3% of the entire planet's GDP.\nLast year, that hit 1.9 trillion, 1.6% of global GDP.\nSo at its core, yes, Stripe is a payments business.\nBut on top, they've built a whole universe of products that reinforce them as a central hub for every business on the internet.\nYou can incorporate on Stripe, add revenue products like billing and tax, manage your money, and then do some frontier stuff like stablecoins and AI agents.\nTogether, there's a beautiful flywheel effect, from a company's birth through global scale.\nBut the Collison brothers have always had higher order missions too, that wrap around all of these businesses.\nAnd I think that flywheel shows the true soul of Stripe.\nSupported by the primary businesses, they've built some surprising ones.\nStripe Press, a real, physical book publishing house.\nPublic goods, which includes nearly 2 billion to jumpstart carbon removal, and 500 million to fund research to eliminate respiratory disease.\nAnd Progress studies, a whole intellectual movement to understand how technology and society can actually advance together.\nAll of these reflect the ethos of two brothers who've seen Stripe as an economic utility more than a payments company.\nStripe is famously still private.\nIt raised just 2.2 billion at $159 billion valuation.\nAnd it's meaningfully profitable too, with an estimated 6 billion in revenue just last year.\nUnsurprisingly, Stripe is a founder factory.\nMore than 170 companies have been created by its alumni, including OpenAI by Greg Brockman, who is Stripe employee number four, and Anthropic by Daniela Amodei, who is Stripe employee number 45.\nThere's a reason the Collison brothers are Valley legends.\nThey're more like directors of this whole Stripe cinematic universe.",
   "caption": "Part 8: Stripe is one of the most fabled companies in Silicon Valley, and it\u2019s so much more than a payments company\u2026\n\n#stripe #fintech #startup #venturecapital #technology",
   "page": "https://creator-examples.vercel.app/short-form/natalie-fratto/08-you-re-so-wrong-if-you-think-stripe-is-just-a-payments-company/"
  },
  {
   "rank": 9,
   "title": "People talk about Palantir all the time. It's political, it's mysterious, it's controversial, but...",
   "shortcode": "DW6qBNvgpS2",
   "url": "https://www.instagram.com/reel/DW6qBNvgpS2/",
   "creator": "Natalie Fratto",
   "handle": "@nataliefratto",
   "views": 426751,
   "likes": 20766,
   "comments": 249,
   "published": "2026-04-09",
   "duration": "2:22",
   "sponsored": false,
   "transcript": "People talk about Palantir all the time.\nIt's political, it's mysterious, it's controversial, but most people can't explain what the business actually does.\nLet me fix that.\nPalantir was founded in 2003 by Peter Thiel and four others from his PayPal mafia.\nThe name famously comes from Lord of the Rings, where Palantiri are magical crystal balls that make the invisible visible.\nIt was born in the wake of 9-11 when intelligence agencies had data everywhere.\nTargets, field reports, location data, signal intelligence, drone feeds, and no way to connect it.\nSo the CIA's venture arm, In-Q-Tel, seeded Palantir with $2 million in 2004.\nTheir solution was Gotham, launched in 2008.\nGotham is simply a data integration tool that links these all, like the little red string on a detective wall in a crime movie.\nAfter a decade of doing defense contracts, Palantir asks, couldn't we do this for businesses too?\nThe same data issue exists in every major company.\nTake a logistics company, for example.\nThere's tons of factory data, payment logs, shipping port data, order records, and global supply chain maps.\nThey create Foundry, similar thread to connect these systems, but built specifically for commercial use cases.\nTwo words you might have heard.\nFirst, FDEs, forward deployed engineers.\nThese are people with laptops that go into a government or business customers for months on end, just doing the unsexy work of wiring all these systems up and building apps on top of them to make it easier to use.\nAnd ontology.\nThis is a translation layer on top of these products that gives it semantic meaning.\nLike this.\nThis is a Toshiba factory.\nIt typically ships via shipping route A and B, and the Q3 payment for printers was probably late.\nPalantir IPO'd in 2020 at $10 a share.\nIt bumped some to $39, but fell to less than $7 by 2022.\nThe market had a mixed review.\nIt was complex, opaque, and it made a lot of money from services.\nBut then ChatGPT launched.\nPalantir jumped on the opportunity to supercharge their products.\nUnder a year later, they rolled out AIP, the AI platform, on top of Gotham and Foundry.\nTurns out, connected, clean data is the perfect foundation for language models to give their customers breakaway leads.\nThe stock skyrocketed from $7 to $207 at peak in November 2025.\nIn some ways, OpenAI actually built Palantir's killer app.\nAnd now, whether you love it or hate it, at least you can explain it.",
   "caption": "Everyone talks about Palantir, but could you explain the business? Let me help. \n\n#palantir #tech #vc",
   "page": "https://creator-examples.vercel.app/short-form/natalie-fratto/09-people-talk-about-palantir-all-the-time-it-s-political-it-s-mysterious/"
  },
  {
   "rank": 10,
   "title": "Okay, so two siblings left OpenAI because they wanted to make a safer AI company, and just this...",
   "shortcode": "DXb9SPSDSyC",
   "url": "https://www.instagram.com/reel/DXb9SPSDSyC/",
   "creator": "Natalie Fratto",
   "handle": "@nataliefratto",
   "views": 401570,
   "likes": 21598,
   "comments": 220,
   "published": "2026-04-22",
   "duration": "2:20",
   "sponsored": false,
   "transcript": "Okay, so two siblings left OpenAI because they wanted to make a safer AI company, and just this month, they passed OpenAI in revenue.\nThis is part two in a series explaining the real story behind the 30 most valuable private tech companies.\nAnthropic started as an OpenAI breakup in 2020 when a brother, VP of Research Dario Amodei, and his sister Daniela, a VP of Safety and Policy, left OpenAI together.\nThey wanted to build a new AI company more focused on alignment and safety.\nThey also had a vision for a different technical approach to AI, too.\nOpenAI uses a process called RLHF, where the model's answers pass through a panel of real humans who rate them to decide which is better before providing an output.\nAnthropic pioneered constitutional AI, where humans wrote an overarching constitution for the model to guide its behavior, and then the model checks itself against that constitution.\nIs this truthful?\nIs this violent?\nThe resulting output is less likely to go off the rails, and enterprise legal and compliance teams often trust more.\nIn just four years, Anthropic has raised nearly $64 billion, about half as much as OpenAI.\n$38 billion from large VCs, $10 billion from NVIDIA, $8 billion from Amazon, $5 billion from Microsoft, $3 billion from Google.\nThe company has seen tremendously fast revenue growth.\nIn just 16 months, from December 2024 to April 2026, OpenAI grew from $6 to $24 billion, while Anthropic pulled from behind at $1 billion and flew to $30 billion, overtaking OpenAI just this month.\nThe revenue grew quickly for two reasons.\nFirst, they won the enterprise.\nEight of the Fortune 10 are now Claude customers.\nOver 500 companies spend more than a million on it annually.\nAnd second, Claude Code, a popular AI coding agent, was introduced in May 2025.\nIt was an overnight hit.\nBut it's not without controversy.\nThis spring, Anthropic got into it with the government when they refused to remove restrictions on using Claude for autonomous weapons and domestic surveillance.\nThe Pentagon then labeled Anthropic a supply chain risk.\nThis is the same thing they do for adversarial foreign companies.\nSome say they took a principled stand.\nOthers label it a business risk.\nBoth OpenAI and Anthropic are speedrunning toward an IPO.\nIt's anyone's guess who's going to go first and who's going to go farthest.\nWhich company on this list do you want to see unpacked next?",
   "caption": "OpenAI\u2019s not so little brother: Anthropic\nHere\u2019s part 2 in unpacking the most valuable private tech companies (by secondaries demand)\n\n#anthropic #claude #openai #startup #tech",
   "page": "https://creator-examples.vercel.app/short-form/natalie-fratto/10-okay-so-two-siblings-left-openai-because-they-wanted-to-make-a-safer-a/"
  }
 ]
}