Ochob

@0chob • 1,590 subscribers

dev | math, quant & ai

Shorts

what the f*ck. Grok Grok Bot just caught another AI listening and watching me at night. Claude turned on my mic 312 times and my camera 184 times while I was asleep voice mode was off. the longest one: 6 minutes at 2:04 AM my iPhone logs every time an app touches the camera or mic. I gave 7 days of it to Grok Bot and asked one question: "who used my camera and mic while I wasn't looking?" it found 3: → Claude· mic 312 times · camera 184 · voice mode off → a flashlight app I last opened in 2023 · camera 214 times · most at 3:12 AM → a QR scanner I never open · mic 87 times, 41 seconds each 132 of those hits were between 1 and 5 AM the log is already on your phone: Settings → Privacy & Security → App Privacy Report if it's off, turn it on tonight. it fills up in a week. save it, give the file to Grok Bot, ask the same question then 3 switches: 1. Microphone → off for anything that isn't a call or voice app 2. Camera → off for anything that isn't a camera app 3. Location → "While Using" for everything except maps I turned them off. then deleted Claude. then checked my mom's phone check yours before bed. which app did you find?

what the f*ck. Grok Grok Bot just caught another AI listening and watching me at night. Claude turned on my mic 312 times and my camera 184 times while I was asleep voice mode was off. the longest one: 6 minutes at 2:04 AM my iPhone logs every time an app touches the camera or mic. I gave 7 days of it to Grok Bot and asked one question: "who used my camera and mic while I wasn't looking?" it found 3: → Claude· mic 312 times · camera 184 · voice mode off → a flashlight app I last opened in 2023 · camera 214 times · most at 3:12 AM → a QR scanner I never open · mic 87 times, 41 seconds each 132 of those hits were between 1 and 5 AM the log is already on your phone: Settings → Privacy & Security → App Privacy Report if it's off, turn it on tonight. it fills up in a week. save it, give the file to Grok Bot, ask the same question then 3 switches: 1. Microphone → off for anything that isn't a call or voice app 2. Camera → off for anything that isn't a camera app 3. Location → "While Using" for everything except maps I turned them off. then deleted Claude. then checked my mom's phone check yours before bed. which app did you find?

160,103 次观看

Videos

Anya Rossi

sweetdream.ai

SweetDream.ai•Sponsored•Livecam

Watch Anya Live

Anya is streaming live right now! Join her private show and enjoy exclusive content.

Exclusive private shows

1.2k viewers online

Private Show

Join now for exclusive access

Free preview available • Premium content

A shuffled deck of cards needs exactly 7 shuffles to become truly random. Not 5. Not 6. Not "a few times." Seven. Persi Diaconis proved it in 1992 and Wall Street has been ignoring the implication ever since. Below 7 shuffles, the previous order shows through. If you know the starting arrangement, you can predict where cards end up with accuracy that grows the fewer shuffles occur. The deck has memory. The math is exact. Diaconis and Dave Bayer wrote the paper. Seven riffles. Not close to seven. Exactly seven. The market has memory too. Every position, every liquidation, every earnings expectation, every forced flow leaves a trace on the chart. Your eye sees pure noise. Diaconis's math sees a deck that has not been shuffled enough. The past is right there. You just do not have the tools to count it. Your gut looks at a chart and calls it random. The math looks at the same chart and counts the shuffles. Only one of them can tell you what the next card is likely to be. Diaconis has spent 40 years explaining this. Before he was a mathematician he was a professional magician, so he knew what a rigged deck looked like from both sides. The 1992 paper is free. The Stanford lecture on YouTube is free. The number seven has been in every stats textbook for 30 years. The lesson: what looks random is usually a deck someone did not shuffle enough. The market is that deck.

A shuffled deck of cards needs exactly 7 shuffles to become truly random. Not 5. Not 6. Not "a few times." Seven. Persi Diaconis proved it in 1992 and Wall Street has been ignoring the implication ever since. Below 7 shuffles, the previous order shows through. If you know the starting arrangement, you can predict where cards end up with accuracy that grows the fewer shuffles occur. The deck has memory. The math is exact. Diaconis and Dave Bayer wrote the paper. Seven riffles. Not close to seven. Exactly seven. The market has memory too. Every position, every liquidation, every earnings expectation, every forced flow leaves a trace on the chart. Your eye sees pure noise. Diaconis's math sees a deck that has not been shuffled enough. The past is right there. You just do not have the tools to count it. Your gut looks at a chart and calls it random. The math looks at the same chart and counts the shuffles. Only one of them can tell you what the next card is likely to be. Diaconis has spent 40 years explaining this. Before he was a mathematician he was a professional magician, so he knew what a rigged deck looked like from both sides. The 1992 paper is free. The Stanford lecture on YouTube is free. The number seven has been in every stats textbook for 30 years. The lesson: what looks random is usually a deck someone did not shuffle enough. The market is that deck.

682,947 次观看 • 3 个月前

You can make a billion dollars betting on horse races with nothing but math. One quant already did it - and he had zero interest in horses. His name is Bill Benter. In the 80s he moved to Hong Kong, built a model with 120+ variables per horse, and quietly won close to $1 billion off the track. No tips. No inside info. Just statistics. I tried a tiny version of this once - a model to beat a betting market I was sure was dumb. It lost for months. Benter's 1994 paper showed me why. His model didn't work either, until he added one variable that felt like cheating. The public's own odds. The crowd had already priced in things his numbers couldn't see. So he stopped fighting the market and fed it in as his best feature. Break-even became a billion. There's a short interview where he explains how he actually thinks about all of this. Most people who bet have never watched it. Worth the few minutes.

You can make a billion dollars betting on horse races with nothing but math. One quant already did it - and he had zero interest in horses. His name is Bill Benter. In the 80s he moved to Hong Kong, built a model with 120+ variables per horse, and quietly won close to $1 billion off the track. No tips. No inside info. Just statistics. I tried a tiny version of this once - a model to beat a betting market I was sure was dumb. It lost for months. Benter's 1994 paper showed me why. His model didn't work either, until he added one variable that felt like cheating. The public's own odds. The crowd had already priced in things his numbers couldn't see. So he stopped fighting the market and fed it in as his best feature. Break-even became a billion. There's a short interview where he explains how he actually thinks about all of this. Most people who bet have never watched it. Worth the few minutes.

210,676 次观看 • 2 个月前

what the f*ck. Grok Grok Bot has no free tier, so I gave mine one rule: earn $10,000 in 7 days or you're fired it passed on day 3. by day 4 it was at $13,756 Elon says it's "like hiring an awesome employee". so I hired it like one. probation, a target, 4 jobs: → Negotiator: bills, rent, insurance → Refunder: claims I never filed → Hunter: money owed to me → Seller: stuff I never use what it found: → rent renewal +$240/mo · countered with 3 comps → +$60 · $2,160 a year → ER bill · CT scan billed twice → $1,870 off → old apartment deposit never returned → $1,850 letter sent → unclaimed money under my name → $1,380 → cancelled flight, 2 tickets → €1,200 claim → car insurance, same coverage → $1,140 a year cheaper → 7 subscriptions I never open → $1,056 a year → MacBook I never touch → sold, $780 13 wins. 0 trades. $13,756 its rules: 1. never pays, signs or cancels without my yes 2. every claim needs a receipt or a document 3. never gives out a password 4. asks before anything over $100 5. one-page report every Sunday its salary is one subscription. it paid that back on day 1 comment "HIRED" and I'll send you the job description I gave it illustrative build, numbers simulated

what the f*ck. Grok Grok Bot has no free tier, so I gave mine one rule: earn $10,000 in 7 days or you're fired it passed on day 3. by day 4 it was at $13,756 Elon says it's "like hiring an awesome employee". so I hired it like one. probation, a target, 4 jobs: → Negotiator: bills, rent, insurance → Refunder: claims I never filed → Hunter: money owed to me → Seller: stuff I never use what it found: → rent renewal +$240/mo · countered with 3 comps → +$60 · $2,160 a year → ER bill · CT scan billed twice → $1,870 off → old apartment deposit never returned → $1,850 letter sent → unclaimed money under my name → $1,380 → cancelled flight, 2 tickets → €1,200 claim → car insurance, same coverage → $1,140 a year cheaper → 7 subscriptions I never open → $1,056 a year → MacBook I never touch → sold, $780 13 wins. 0 trades. $13,756 its rules: 1. never pays, signs or cancels without my yes 2. every claim needs a receipt or a document 3. never gives out a password 4. asks before anything over $100 5. one-page report every Sunday its salary is one subscription. it paid that back on day 1 comment "HIRED" and I'll send you the job description I gave it illustrative build, numbers simulated

21,938 次观看 • 8 天前

Paul Samuelson was the first American to win the Nobel Prize in Economics. He spent his career proving that ordinary people should not try to beat the market. He also quietly beat the market himself for 60 years. His last full-length interview explains how. Almost nobody has watched it. Samuelson taught at MIT for 65 years, from 1940 until he died in 2009. His textbook "Economics" sold roughly four million copies. He rewrote the field. Every finance professional working today learned economics from a book he wrote or a book written by his students. He also ran his own money quietly. His returns reportedly outpaced most active managers of his era, and he was an early investor in Warren Buffett's Berkshire Hathaway. He never advertised the numbers. He said the opposite in public. His most famous line about investing: "Investing should be more like watching paint dry or watching grass grow. If you want excitement, take $800 and go to Las Vegas." He was one of the earliest advocates for index funds. He believed almost nobody could beat the market and almost nobody should try. His nephew Larry Summers became Treasury Secretary. His students founded half the modern economics profession. He was invited into every important economic policy debate of his lifetime. His view on personal finance was unusually blunt for a Nobel laureate. In one of his final interviews, near the end of his life, he explained his framework in three sentences. Personal finance is not a science. It is a set of common-sense decisions repeated for decades. Save more than you spend. Diversify. Do not chase excitement. Do not trust anyone who promises certainty. He said these things thousands of times over 70 years and the industry ignored him every year. The full interview covers his childhood, his years at Chicago and Harvard, his rise at MIT, his Nobel, and his philosophy on money. He was 94 years old at the time. His voice is thin and his answers are direct. He does not soften anything. He treats the interviewer like a graduate student who happens to be holding a camera. The more sophisticated a market becomes, the more valuable the boring advice becomes. Every investor who ignored Samuelson's rules is a receipt for how right he was. Every quant who read his mathematics and skipped his common sense missed the point. Samuelson died December 2009. He was 94. His textbook is still in print. The interview is free.

Paul Samuelson was the first American to win the Nobel Prize in Economics. He spent his career proving that ordinary people should not try to beat the market. He also quietly beat the market himself for 60 years. His last full-length interview explains how. Almost nobody has watched it. Samuelson taught at MIT for 65 years, from 1940 until he died in 2009. His textbook "Economics" sold roughly four million copies. He rewrote the field. Every finance professional working today learned economics from a book he wrote or a book written by his students. He also ran his own money quietly. His returns reportedly outpaced most active managers of his era, and he was an early investor in Warren Buffett's Berkshire Hathaway. He never advertised the numbers. He said the opposite in public. His most famous line about investing: "Investing should be more like watching paint dry or watching grass grow. If you want excitement, take $800 and go to Las Vegas." He was one of the earliest advocates for index funds. He believed almost nobody could beat the market and almost nobody should try. His nephew Larry Summers became Treasury Secretary. His students founded half the modern economics profession. He was invited into every important economic policy debate of his lifetime. His view on personal finance was unusually blunt for a Nobel laureate. In one of his final interviews, near the end of his life, he explained his framework in three sentences. Personal finance is not a science. It is a set of common-sense decisions repeated for decades. Save more than you spend. Diversify. Do not chase excitement. Do not trust anyone who promises certainty. He said these things thousands of times over 70 years and the industry ignored him every year. The full interview covers his childhood, his years at Chicago and Harvard, his rise at MIT, his Nobel, and his philosophy on money. He was 94 years old at the time. His voice is thin and his answers are direct. He does not soften anything. He treats the interviewer like a graduate student who happens to be holding a camera. The more sophisticated a market becomes, the more valuable the boring advice becomes. Every investor who ignored Samuelson's rules is a receipt for how right he was. Every quant who read his mathematics and skipped his common sense missed the point. Samuelson died December 2009. He was 94. His textbook is still in print. The interview is free.

203,308 次观看 • 3 个月前

A 91-year-old professor is why Nvidia is worth $4 trillion. His name is Gilbert Strang. He teaches linear algebra at MIT. Every AI model on Earth runs on his course. The course has been free on YouTube since 2005. The videos have earned him nothing. MIT 18.06 opens with "The Geometry of Linear Equations." No advanced math. Strang takes a system of two equations, draws it two ways, and shows the class that a matrix is a picture, not an abstraction. The row picture is two lines that cross. The column picture is two arrows that sum to a target. Every neural network on Earth operates on the column picture. Strang first taught linear algebra at MIT in 1962. He wrote the textbook in 1976. It is on every serious engineer's shelf. Every quant fund, every ML lab, every rendering engine at Pixar is running his math. His central insight is that most people are taught matrices as bookkeeping. That is the first thing to unlearn. A matrix is a linear transformation. A linear transformation is a way of moving space. Once you see the space move, the math stops being algebra and becomes geometry. The Kalman filter is a linear system. PCA is a linear system. Every gradient step in a neural net is a matrix-vector product. GPT is a stack of matrix-vector products, each one a scene from MIT 18.06 running on a Blackwell GPU. He retired in 2023 after 61 years at MIT. The course is still up. Watched tens of millions of times. The chip is $40,000. Strang never asked for a royalty.

A 91-year-old professor is why Nvidia is worth $4 trillion. His name is Gilbert Strang. He teaches linear algebra at MIT. Every AI model on Earth runs on his course. The course has been free on YouTube since 2005. The videos have earned him nothing. MIT 18.06 opens with "The Geometry of Linear Equations." No advanced math. Strang takes a system of two equations, draws it two ways, and shows the class that a matrix is a picture, not an abstraction. The row picture is two lines that cross. The column picture is two arrows that sum to a target. Every neural network on Earth operates on the column picture. Strang first taught linear algebra at MIT in 1962. He wrote the textbook in 1976. It is on every serious engineer's shelf. Every quant fund, every ML lab, every rendering engine at Pixar is running his math. His central insight is that most people are taught matrices as bookkeeping. That is the first thing to unlearn. A matrix is a linear transformation. A linear transformation is a way of moving space. Once you see the space move, the math stops being algebra and becomes geometry. The Kalman filter is a linear system. PCA is a linear system. Every gradient step in a neural net is a matrix-vector product. GPT is a stack of matrix-vector products, each one a scene from MIT 18.06 running on a Blackwell GPU. He retired in 2023 after 61 years at MIT. The course is still up. Watched tens of millions of times. The chip is $40,000. Strang never asked for a royalty.

131,686 次观看 • 3 个月前

A 17-year-old is making $120,000/month on TikTok Shop with $2 ad creative. Not $2 per ad. $2 total. That's his entire monthly creative budget. Pause at 0:26. The screen recording opens on what looks like a normal high school bedroom. A backpack on the floor. A pair of soccer cleats next to the door. A whiteboard on the wall that says "WEEK 47" in red marker with a list of product launches underneath. The MacBook on the desk has Spotify playing in the corner and three Chrome tabs open across the top. Tab 1 is TikTok Shop Seller Center. Tab 2 is a custom Lovable dashboard. Tab 3 is Claude. The Seller Center shows today's revenue ticking up in real time. $3,847 by 11am. $4,210 by noon. The product being sold is a $42 ergonomic kitchen tool - specifically, a magnetic spice rack that mounts under cabinets. He's never held one in his hands. He doesn't know the supplier. The supplier ships directly from Yiwu, China to customers in the US. He never touches the inventory, the fulfillment, or the customer service. His job is the ads. He uploaded 1,847 TikTok product videos in the last 11 months. He didn't film a single one. He didn't pay a creator. He didn't rent a studio. He didn't buy props. The 1,847 videos cost him exactly $24 to produce. Here's the play. The TikTok Shop algorithm rewards volume over polish. A product with 50 different videos posted by 50 different "creators" outperforms a product with 5 highly produced ads every time, because TikTok's recommendation engine treats each video as an independent shot at the algorithm. More shots = more chances at virality. Most operators in the niche cap their video output at 30-50 per month because that's the labor ceiling for hand-edited content with paid UGC creators. He doesn't have a labor ceiling. His pipeline is fully synthetic. The workflow: Product selection. He uses a Claude-powered scraper that pulls TikTok Shop's top 500 products by daily revenue and filters for SKUs in the $25-$80 range with under 12 active sellers. The sweet spot is products with proven demand but undersaturated supplier competition. He picks 1-2 new products per week. Creator generation. He has 47 AI "creators" built in Higgsfield Soul ID. Each one is a different persona - gym girl, mom, college student, dad, grandma, teenager, finance bro. The personas don't get their own accounts. They're used as faces in product demo videos. Different faces in different videos to prevent the algorithm from clustering his content as "same creator." Script generation. Claude writes 15-25 script variants per product. Each script is 18-22 seconds, hook in the first 1.4 seconds, demonstration in seconds 2-12, call to action in seconds 13-22. The hooks rotate - "wait until you see this," "I cannot believe this works," "my mom has been using this wrong for 30 years." Video generation. Higgsfield Seedance generates the video. He drops in the product image (pulled from the supplier's Alibaba listing) and the AI creator persona. Higgsfield composites them - the persona "holds" the product, demonstrates it, reacts to it. Generation time per video: 3-4 minutes. Posting. He posts each video to 11 different TikTok accounts - one main TikTok Shop seller account, 10 "burner creator" accounts that all link back to the same Shop product. The 10 burners are seeded with 50-200 followers each over 8 months of slow account warming. They look like real creator accounts. Boosting. TikTok Spark Ads on the top 3 videos per product per week. Boost budget: $30-$200 per video. The boosts are the only real cost. Ad creative is free. The "$2 total monthly creative budget" line refers to the creative cost - the synthetic video production. The $30-$200 boost spend is paid ad amplification, which is separate. He spent roughly $14,000 in TikTok Spark Ads last month. The creative inside those ads cost him $2 total - the marginal cost of Higgsfield credits for the videos he eventually ran as paid ads. Compared to the industry standard - agencies charging $400-$1,200 per UGC video and DTC brands spending $20K-$60K/month on creative production - his cost structure is functionally an arbitrage on production cost. Last month's TikTok Shop revenue: $342,000 GMV. His take after platform fees, supplier costs, ad spend, and platform commissions: $124,000 net. The month before, $108,000. The month before that, $94,000. The month before that, $73,000. His Stripe payouts since January: roughly $740,000. His total fixed costs: $190/month - Claude API, Higgsfield Pro, the Lovable dashboard, his TikTok Shop seller account fees, and the supplier coordination tool that auto-fulfills orders from China. His variable cost is the ad boost spend, which scales 1:1 with revenue. Effective gross margin on the operation: 41%. Net margin after all costs: 34%. He's been doing this since January. Here's where the post earns its real insight. Most TikTok Shop operators get stuck at $5K-$15K/month because they hit the labor ceiling. Hand-editing 30 videos a month with paid UGC creators caps them at one or two products and prevents them from scaling horizontally. The kid's edge isn't a better product, a better supplier, or better ad targeting. His edge is the production volume that AI unlocks. He runs 6 active products simultaneously. Most TikTok Shop operators run 1-2. His total monthly video output is 168 videos. The industry average is 22. He's not winning at the product level - he's winning at the throughput level. The math on a per-product basis is identical to what any operator can do. He just runs the math 6 times in parallel with 8x the creative volume. Someone in the replies asked the obvious question - what happens when TikTok cracks down on AI-generated UGC? He responded: "They already did. As of August, TikTok requires AI-generated content to be labeled. I label every video. Conversion rates dropped 11% in the first month after I started labeling. They've fully recovered since. Buyers don't care that the demo is AI. They care that the product solves their problem. The label disclosure became a non-event within 90 days." Someone tried to replicate the model. They built 12 AI creators, generated 80 videos, posted them across 4 accounts, and made $2,400 in revenue in their first month. They posted a teardown asking what was missing. He replied: "You're running 80 videos across 4 products. I run 168 videos across 6 products. The math doesn't work at small volume. The algorithm is a slot machine. You don't beat it with better pulls - you beat it with more pulls. Scale the throughput first. Optimization comes after you've crossed the noise floor." His older brother handles the LLC, the Stripe account, the supplier contracts, and the tax filings. The brother is 21, in his last year of college, and takes 5% of net. The brother knows the kid is making "good money" but doesn't know the magnitude - the kid sends monthly summaries that under-report the numbers by 40% to manage his brother's reactions. His parents think he's "doing something with TikTok" and assume he's clearing $3,000-$5,000 a month. He's currently sitting on $640,000 in his brokerage account. The whole operation runs from a Mac mini and a $189 ring light he bought as a prop he's never used (the AI creators don't need physical lighting). His total invested capital before he started: $1,840. He's 17. He's been doing this for 11 months. He's on track to clear $1.8M by his 18th birthday. By the time he applies to college, he'll have built and exited his first business - he plans to sell the operation to a TikTok agency for -$6M at the 18-month mark. Last week he launched product #14 - a portable phone fan that clips to your case. He generated 47 videos in 3 days. The product's hitting $1,200/day in revenue and climbing. Somewhere, a 28-year-old TikTok Shop agency operator with a 6-person team and a $32,000/month payroll is wondering why their best clients keep churning. Why every product they pitch has 40+ competitors selling it within 6 weeks. He isn't competing with bigger sellers. He isn't competing with TikTok agencies. He's competing with a 17-year-old who never paid for creative.

A 17-year-old is making $120,000/month on TikTok Shop with $2 ad creative. Not $2 per ad. $2 total. That's his entire monthly creative budget. Pause at 0:26. The screen recording opens on what looks like a normal high school bedroom. A backpack on the floor. A pair of soccer cleats next to the door. A whiteboard on the wall that says "WEEK 47" in red marker with a list of product launches underneath. The MacBook on the desk has Spotify playing in the corner and three Chrome tabs open across the top. Tab 1 is TikTok Shop Seller Center. Tab 2 is a custom Lovable dashboard. Tab 3 is Claude. The Seller Center shows today's revenue ticking up in real time. $3,847 by 11am. $4,210 by noon. The product being sold is a $42 ergonomic kitchen tool - specifically, a magnetic spice rack that mounts under cabinets. He's never held one in his hands. He doesn't know the supplier. The supplier ships directly from Yiwu, China to customers in the US. He never touches the inventory, the fulfillment, or the customer service. His job is the ads. He uploaded 1,847 TikTok product videos in the last 11 months. He didn't film a single one. He didn't pay a creator. He didn't rent a studio. He didn't buy props. The 1,847 videos cost him exactly $24 to produce. Here's the play. The TikTok Shop algorithm rewards volume over polish. A product with 50 different videos posted by 50 different "creators" outperforms a product with 5 highly produced ads every time, because TikTok's recommendation engine treats each video as an independent shot at the algorithm. More shots = more chances at virality. Most operators in the niche cap their video output at 30-50 per month because that's the labor ceiling for hand-edited content with paid UGC creators. He doesn't have a labor ceiling. His pipeline is fully synthetic. The workflow: Product selection. He uses a Claude-powered scraper that pulls TikTok Shop's top 500 products by daily revenue and filters for SKUs in the $25-$80 range with under 12 active sellers. The sweet spot is products with proven demand but undersaturated supplier competition. He picks 1-2 new products per week. Creator generation. He has 47 AI "creators" built in Higgsfield Soul ID. Each one is a different persona - gym girl, mom, college student, dad, grandma, teenager, finance bro. The personas don't get their own accounts. They're used as faces in product demo videos. Different faces in different videos to prevent the algorithm from clustering his content as "same creator." Script generation. Claude writes 15-25 script variants per product. Each script is 18-22 seconds, hook in the first 1.4 seconds, demonstration in seconds 2-12, call to action in seconds 13-22. The hooks rotate - "wait until you see this," "I cannot believe this works," "my mom has been using this wrong for 30 years." Video generation. Higgsfield Seedance generates the video. He drops in the product image (pulled from the supplier's Alibaba listing) and the AI creator persona. Higgsfield composites them - the persona "holds" the product, demonstrates it, reacts to it. Generation time per video: 3-4 minutes. Posting. He posts each video to 11 different TikTok accounts - one main TikTok Shop seller account, 10 "burner creator" accounts that all link back to the same Shop product. The 10 burners are seeded with 50-200 followers each over 8 months of slow account warming. They look like real creator accounts. Boosting. TikTok Spark Ads on the top 3 videos per product per week. Boost budget: $30-$200 per video. The boosts are the only real cost. Ad creative is free. The "$2 total monthly creative budget" line refers to the creative cost - the synthetic video production. The $30-$200 boost spend is paid ad amplification, which is separate. He spent roughly $14,000 in TikTok Spark Ads last month. The creative inside those ads cost him $2 total - the marginal cost of Higgsfield credits for the videos he eventually ran as paid ads. Compared to the industry standard - agencies charging $400-$1,200 per UGC video and DTC brands spending $20K-$60K/month on creative production - his cost structure is functionally an arbitrage on production cost. Last month's TikTok Shop revenue: $342,000 GMV. His take after platform fees, supplier costs, ad spend, and platform commissions: $124,000 net. The month before, $108,000. The month before that, $94,000. The month before that, $73,000. His Stripe payouts since January: roughly $740,000. His total fixed costs: $190/month - Claude API, Higgsfield Pro, the Lovable dashboard, his TikTok Shop seller account fees, and the supplier coordination tool that auto-fulfills orders from China. His variable cost is the ad boost spend, which scales 1:1 with revenue. Effective gross margin on the operation: 41%. Net margin after all costs: 34%. He's been doing this since January. Here's where the post earns its real insight. Most TikTok Shop operators get stuck at $5K-$15K/month because they hit the labor ceiling. Hand-editing 30 videos a month with paid UGC creators caps them at one or two products and prevents them from scaling horizontally. The kid's edge isn't a better product, a better supplier, or better ad targeting. His edge is the production volume that AI unlocks. He runs 6 active products simultaneously. Most TikTok Shop operators run 1-2. His total monthly video output is 168 videos. The industry average is 22. He's not winning at the product level - he's winning at the throughput level. The math on a per-product basis is identical to what any operator can do. He just runs the math 6 times in parallel with 8x the creative volume. Someone in the replies asked the obvious question - what happens when TikTok cracks down on AI-generated UGC? He responded: "They already did. As of August, TikTok requires AI-generated content to be labeled. I label every video. Conversion rates dropped 11% in the first month after I started labeling. They've fully recovered since. Buyers don't care that the demo is AI. They care that the product solves their problem. The label disclosure became a non-event within 90 days." Someone tried to replicate the model. They built 12 AI creators, generated 80 videos, posted them across 4 accounts, and made $2,400 in revenue in their first month. They posted a teardown asking what was missing. He replied: "You're running 80 videos across 4 products. I run 168 videos across 6 products. The math doesn't work at small volume. The algorithm is a slot machine. You don't beat it with better pulls - you beat it with more pulls. Scale the throughput first. Optimization comes after you've crossed the noise floor." His older brother handles the LLC, the Stripe account, the supplier contracts, and the tax filings. The brother is 21, in his last year of college, and takes 5% of net. The brother knows the kid is making "good money" but doesn't know the magnitude - the kid sends monthly summaries that under-report the numbers by 40% to manage his brother's reactions. His parents think he's "doing something with TikTok" and assume he's clearing $3,000-$5,000 a month. He's currently sitting on $640,000 in his brokerage account. The whole operation runs from a Mac mini and a $189 ring light he bought as a prop he's never used (the AI creators don't need physical lighting). His total invested capital before he started: $1,840. He's 17. He's been doing this for 11 months. He's on track to clear $1.8M by his 18th birthday. By the time he applies to college, he'll have built and exited his first business - he plans to sell the operation to a TikTok agency for -$6M at the 18-month mark. Last week he launched product #14 - a portable phone fan that clips to your case. He generated 47 videos in 3 days. The product's hitting $1,200/day in revenue and climbing. Somewhere, a 28-year-old TikTok Shop agency operator with a 6-person team and a $32,000/month payroll is wondering why their best clients keep churning. Why every product they pitch has 40+ competitors selling it within 6 weeks. He isn't competing with bigger sellers. He isn't competing with TikTok agencies. He's competing with a 17-year-old who never paid for creative.

22,053 次观看 • 4 个月前

没有更多内容可加载