Video wird geladen...

Video konnte nicht geladen werden

Zur Startseite

a Renaissance intern posted his model on GitHub for 6 hours before deleting it someone in a Discord dropped the link at 1am. "this looks real. star count is zero. repo is 4 hours old. readme has internal formatting" i clicked. it was a full prediction market scoring model....

hanakoxbt's profile picture

Hanako

10,144 subscribers

117,572 Aufrufe • vor 2 Monaten •via X (Twitter)

BildungWissenschaft & TechnologieFinanzen
LIVE

Anya Rossi• Live Now

Private livecam show

0 Kommentare

Keine Kommentare verfügbar

Kommentare vom Original-Post werden hier angezeigt

Ähnliche Videos

an ex-OpenAI researcher pushed his trading system to GitHub the night he got fired someone in a crypto Discord dropped the link at 2am. "this repo appeared an hour ago. zero stars. readme says internal use only. the code looks real" i clicked. it was a complete prediction market execution engine. bayesian consensus scoring. whale divergence tracker. kelly position sizing. edge detection on 5-minute BTC binary markets. clean Python. commented like someone who writes for a team. i ran git clone before finishing the readme. by 4am the repo was gone. deleted. commits. branches. history. everything. but i had it. opened a Chinese AI. gave it the code and one prompt. "adapt this for Polymarket. deploy it live. keep the scoring logic intact" 18,000 stars. open source HFT framework. the AI used it as the execution backbone. order routing. position management. risk controls. merged the stolen scoring model with hummingbot's engine. system was live by sunrise. 288 BTC markets per day. every 5 minutes. the crowd bets direction. his model doesn't care about direction. it buys whichever side the crowd underpriced. pair cost below 99c = profit regardless of outcome. the part that scared me: it worked immediately. first trade. green. second trade. green. third. fourth. fifth. all green. like it was built by someone who actually knew what they were doing. because it was. edge table live: BTC+100K 15m +7.5c. BTC+103K 30m spread 7c. BTC+101K +12.0c. BTC 0xd8dA...6045 +$9,260. 92%. > 0xBE0e...48CD +$7,147. 85%. > 0x47ac...d563 +$5,874. 81%. 323 trades. 85% win rate. sharpe 3.96. drawdown -0.9%. +$19,812 from $1,100 seed. my setup: > AI - $20/mo > VPS - $5/mo > hummingbot - free > his code - free now copy setup you can here: checked his GitHub profile last week. still active. new repos. normal stuff. open source contributions. no mention of the deleted repo anywhere. he has no idea someone cloned it four minutes before it disappeared. and now you're reading what was in it.
0:30

an ex-OpenAI researcher pushed his trading system to GitHub the night he got fired someone in a crypto Discord dropped the link at 2am. "this repo appeared an hour ago. zero stars. readme says internal use only. the code looks real" i clicked. it was a complete prediction market execution engine. bayesian consensus scoring. whale divergence tracker. kelly position sizing. edge detection on 5-minute BTC binary markets. clean Python. commented like someone who writes for a team. i ran git clone before finishing the readme. by 4am the repo was gone. deleted. commits. branches. history. everything. but i had it. opened a Chinese AI. gave it the code and one prompt. "adapt this for Polymarket. deploy it live. keep the scoring logic intact" 18,000 stars. open source HFT framework. the AI used it as the execution backbone. order routing. position management. risk controls. merged the stolen scoring model with hummingbot's engine. system was live by sunrise. 288 BTC markets per day. every 5 minutes. the crowd bets direction. his model doesn't care about direction. it buys whichever side the crowd underpriced. pair cost below 99c = profit regardless of outcome. the part that scared me: it worked immediately. first trade. green. second trade. green. third. fourth. fifth. all green. like it was built by someone who actually knew what they were doing. because it was. edge table live: BTC+100K 15m +7.5c. BTC+103K 30m spread 7c. BTC+101K +12.0c. BTC 0xd8dA...6045 +$9,260. 92%. > 0xBE0e...48CD +$7,147. 85%. > 0x47ac...d563 +$5,874. 81%. 323 trades. 85% win rate. sharpe 3.96. drawdown -0.9%. +$19,812 from $1,100 seed. my setup: > AI - $20/mo > VPS - $5/mo > hummingbot - free > his code - free now copy setup you can here: checked his GitHub profile last week. still active. new repos. normal stuff. open source contributions. no mention of the deleted repo anywhere. he has no idea someone cloned it four minutes before it disappeared. and now you're reading what was in it.

Hanako

23,546 Aufrufe • vor 2 Monaten

a Citadel intern told me something at a party he probably shouldn't have it was on a rooftop in brooklyn. i mentioned i trade prediction markets. he got quiet for a second. "we have a model for that. it scores every contract on four factors. when all four align we enter. when any breaks we exit. that's it" i asked what the four factors are. he looked around. then said it fast like he was confessing. "cross-market divergence. disposition coefficient. capital velocity. pair network correlation" I didn't know what half of that meant. but i memorized it. went home. 11pm. opened Claude. "here are four scoring factors from a quant fund. build a terminal that runs all four on prediction markets" Claude asked one question: "Where's the data?" I sent him one repo: 86 million trades. every wallet. every entry. every outcome three weeks later i'm sitting in my apartment watching a screen i barely understand print money. the disposition meter alone changed everything. it measures how you exit - not how you enter. top wallets capture 86% of winner value and cut losers at 12%. everyone else captures 58% and holds losers to 41%. same exact entries. the exits make it a completely different game. capital velocity: 49x. every dollar gets recycled 49 times before the average trader recycles once. the terminal found 42 pair correlations across 11 markets. when MSFT beats Q3 is priced at 80c but the model reads 93% - it enters. when the gap closes 2 hours later - it exits. no opinions. no news. just four numbers that either align or don't. his fund runs this with a floor of PhDs and $800M AUM. my setup: > Claude - $20/month > VPS - $5/month > poly_data repo - free > Polymarket API - free $25/month. no team. no office. no Bloomberg. 280 trades so far. 70% win rate. $800 seed. four bots splitting the work: pulse_alpha +$299. arb_hunter +$558. trend_rider +$337. cal_engine +$719. +$11,514 total. copytrade here: he texted me last week. "delete everything i told you" too late.
0:38

a Citadel intern told me something at a party he probably shouldn't have it was on a rooftop in brooklyn. i mentioned i trade prediction markets. he got quiet for a second. "we have a model for that. it scores every contract on four factors. when all four align we enter. when any breaks we exit. that's it" i asked what the four factors are. he looked around. then said it fast like he was confessing. "cross-market divergence. disposition coefficient. capital velocity. pair network correlation" I didn't know what half of that meant. but i memorized it. went home. 11pm. opened Claude. "here are four scoring factors from a quant fund. build a terminal that runs all four on prediction markets" Claude asked one question: "Where's the data?" I sent him one repo: 86 million trades. every wallet. every entry. every outcome three weeks later i'm sitting in my apartment watching a screen i barely understand print money. the disposition meter alone changed everything. it measures how you exit - not how you enter. top wallets capture 86% of winner value and cut losers at 12%. everyone else captures 58% and holds losers to 41%. same exact entries. the exits make it a completely different game. capital velocity: 49x. every dollar gets recycled 49 times before the average trader recycles once. the terminal found 42 pair correlations across 11 markets. when MSFT beats Q3 is priced at 80c but the model reads 93% - it enters. when the gap closes 2 hours later - it exits. no opinions. no news. just four numbers that either align or don't. his fund runs this with a floor of PhDs and $800M AUM. my setup: > Claude - $20/month > VPS - $5/month > poly_data repo - free > Polymarket API - free $25/month. no team. no office. no Bloomberg. 280 trades so far. 70% win rate. $800 seed. four bots splitting the work: pulse_alpha +$299. arb_hunter +$558. trend_rider +$337. cal_engine +$719. +$11,514 total. copytrade here: he texted me last week. "delete everything i told you" too late.

Hanako

1,736,545 Aufrufe • vor 2 Monaten

a Goldman Sachs engineer sat next to me on a delayed flight from JFK four hours on the tarmac. nowhere to go. we started talking. i told him i trade prediction markets with Claude. he almost choked on his ginger ale. "you do WHAT with an AI?" i showed him my terminal. he went quiet. scrolled through it for two full minutes. "this is close to what we run internally. how did you get the scoring model?" i told him. one GitHub repo with 2,600 stars. Claude connected the dots. official Polymarket framework. LLM-powered trading agents. open source. free. i gave Claude the repo and said: build me a grid that scans 700+ markets per hour and enters when the edge exceeds 9 cents. he didn't ask twice. the Goldman guy kept scrolling. "you have a disposition meter. a kelly gauge. a correlation matrix. we spent $4M building this infrastructure" i told him my total cost. Claude $20/mo + VPS $5/mo he laughed. not the funny kind. that was three weeks ago. today the terminal shows: +$16,976. 439 trades. 73% win rate. avg hold 7h. > Falcon Heavy - mkt 68c. model 95c. +$716. > MSFT antitrust - mkt 59c. model 77c. +$670. > Blizzard NY - mkt 74c. model 85c. +$772. > BOJ rate hike - mkt 52c. model 60c. +$422. four bots split everything: - pulse_alpha +$299. - arb_hunter +$558. - trend_rider +$337. - cal_engine +$719. kelly fraction 0.118. correlation matrix tracking 6 asset pairs. when one breaks - something is wrong. the grid trades it. copytrade here: he DM'd me yesterday. "i showed your screenshot to my desk. they want to know who built your infra" i said Claude. he said "seriously" i said seriously.
0:31

a Goldman Sachs engineer sat next to me on a delayed flight from JFK four hours on the tarmac. nowhere to go. we started talking. i told him i trade prediction markets with Claude. he almost choked on his ginger ale. "you do WHAT with an AI?" i showed him my terminal. he went quiet. scrolled through it for two full minutes. "this is close to what we run internally. how did you get the scoring model?" i told him. one GitHub repo with 2,600 stars. Claude connected the dots. official Polymarket framework. LLM-powered trading agents. open source. free. i gave Claude the repo and said: build me a grid that scans 700+ markets per hour and enters when the edge exceeds 9 cents. he didn't ask twice. the Goldman guy kept scrolling. "you have a disposition meter. a kelly gauge. a correlation matrix. we spent $4M building this infrastructure" i told him my total cost. Claude $20/mo + VPS $5/mo he laughed. not the funny kind. that was three weeks ago. today the terminal shows: +$16,976. 439 trades. 73% win rate. avg hold 7h. > Falcon Heavy - mkt 68c. model 95c. +$716. > MSFT antitrust - mkt 59c. model 77c. +$670. > Blizzard NY - mkt 74c. model 85c. +$772. > BOJ rate hike - mkt 52c. model 60c. +$422. four bots split everything: - pulse_alpha +$299. - arb_hunter +$558. - trend_rider +$337. - cal_engine +$719. kelly fraction 0.118. correlation matrix tracking 6 asset pairs. when one breaks - something is wrong. the grid trades it. copytrade here: he DM'd me yesterday. "i showed your screenshot to my desk. they want to know who built your infra" i said Claude. he said "seriously" i said seriously.

Hanako

286,427 Aufrufe • vor 2 Monaten

An OpenAI researcher sat down next to me at a coffee shop in Mission District I had my terminal open. Three panels. Live trades scrolling. He was reading something on his laptop. Glanced over. Stopped reading. "That's not a dashboard. That's a live scoring engine. What model is running that" I told him. Claude Code. Four repos. $25 a month. He closed his laptop. "I work at OpenAI. We benchmarked Claude internally last month. You're using it to trade prediction markets?" I opened one link. 86 million trades. Every wallet. Every entry. Every exit. The entire Polymarket history since day one. "This is public? We quoted a seven-figure budget to reconstruct this kind of dataset from on-chain data. The project is still in review" I told him Claude Code connects directly. It reads the whole dataset. Finds the wallets that win. Then finds WHY they win. Then copies the pattern. He pulled his chair closer. "Walk me through the exit logic" Top wallets exit before resolution 91% of the time. They capture 86% of the move and cut losers at 12%. Everyone else holds to 58%. Same entries. Completely different exits. My bot cuts at 85% of expected move. Or on a 3x volume spike. Whichever hits first. "Who gave you that threshold" Claude Code found it in poly_data. In about 20 minutes. "We had a team of nine working on this exact problem for six months. They never shipped it. You did it in a weekend with a competitor's model" I opened another link. Three commands. 500+ markets. No API key. Claude scores them in 20 minutes. "That's our internal eval pipeline. Except it took us six months and you built it on a Saturday" My setup: Claude API - $20/mo VPS - $5/mo poly_data - free polymarket-cli - free 19 days. 4 agents. 74% win rate. +$9,400. Copytrade here: I showed him the article where I broke down every repo, every command, every dollar. He read it for five minutes. Then looked up. "You just published what we presented to Sam last quarter. Using the other team's model" He texted me the next morning. "My director found your thread. Take it down" Too late.
0:20

An OpenAI researcher sat down next to me at a coffee shop in Mission District I had my terminal open. Three panels. Live trades scrolling. He was reading something on his laptop. Glanced over. Stopped reading. "That's not a dashboard. That's a live scoring engine. What model is running that" I told him. Claude Code. Four repos. $25 a month. He closed his laptop. "I work at OpenAI. We benchmarked Claude internally last month. You're using it to trade prediction markets?" I opened one link. 86 million trades. Every wallet. Every entry. Every exit. The entire Polymarket history since day one. "This is public? We quoted a seven-figure budget to reconstruct this kind of dataset from on-chain data. The project is still in review" I told him Claude Code connects directly. It reads the whole dataset. Finds the wallets that win. Then finds WHY they win. Then copies the pattern. He pulled his chair closer. "Walk me through the exit logic" Top wallets exit before resolution 91% of the time. They capture 86% of the move and cut losers at 12%. Everyone else holds to 58%. Same entries. Completely different exits. My bot cuts at 85% of expected move. Or on a 3x volume spike. Whichever hits first. "Who gave you that threshold" Claude Code found it in poly_data. In about 20 minutes. "We had a team of nine working on this exact problem for six months. They never shipped it. You did it in a weekend with a competitor's model" I opened another link. Three commands. 500+ markets. No API key. Claude scores them in 20 minutes. "That's our internal eval pipeline. Except it took us six months and you built it on a Saturday" My setup: Claude API - $20/mo VPS - $5/mo poly_data - free polymarket-cli - free 19 days. 4 agents. 74% win rate. +$9,400. Copytrade here: I showed him the article where I broke down every repo, every command, every dollar. He read it for five minutes. Then looked up. "You just published what we presented to Sam last quarter. Using the other team's model" He texted me the next morning. "My director found your thread. Take it down" Too late.

Lunar

159,590 Aufrufe • vor 2 Monaten

An ex-Anthropic engineer told me something at a party he probably shouldn't have. It was in SF. Someone's rooftop. I mentioned I run trading agents on Claude. He went quiet. "You're doing it wrong. Everyone is" I asked what he meant. "Claude is a runtime. Not a chatbox. You're supposed to pair it with repos" He pulled out his phone. Opened one GitHub link. 14,000 stars. Every workflow pattern they built internally before it went public. Agents. Tool use. Evals. Citations. The entire architecture. "Everyone types prompts. That's not how we use it. You connect Claude to a codebase. It reads. It understands. It builds on top of what's already there" I went home at 2am. Connected Claude Code to poly_data - 86 million Polymarket trades. Every wallet. Every entry. Claude didn't guess. It read the data and built detectors. First week: +$1,400. Second week: +$3,800. Right now: +$9,100. 4 agents. 74% win rate. His team runs this with a floor of PhDs and $800M AUM. My setup: Claude + a VPS. $25/month. The repos are free. Copytrade here: I asked him what separates his firm from everyone else. "Honestly? Keyboard shortcuts and repo structure. That's it. The model is the same for everyone" He texted me two days later. "Delete everything I told you" Too late.
0:21

An ex-Anthropic engineer told me something at a party he probably shouldn't have. It was in SF. Someone's rooftop. I mentioned I run trading agents on Claude. He went quiet. "You're doing it wrong. Everyone is" I asked what he meant. "Claude is a runtime. Not a chatbox. You're supposed to pair it with repos" He pulled out his phone. Opened one GitHub link. 14,000 stars. Every workflow pattern they built internally before it went public. Agents. Tool use. Evals. Citations. The entire architecture. "Everyone types prompts. That's not how we use it. You connect Claude to a codebase. It reads. It understands. It builds on top of what's already there" I went home at 2am. Connected Claude Code to poly_data - 86 million Polymarket trades. Every wallet. Every entry. Claude didn't guess. It read the data and built detectors. First week: +$1,400. Second week: +$3,800. Right now: +$9,100. 4 agents. 74% win rate. His team runs this with a floor of PhDs and $800M AUM. My setup: Claude + a VPS. $25/month. The repos are free. Copytrade here: I asked him what separates his firm from everyone else. "Honestly? Keyboard shortcuts and repo structure. That's it. The model is the same for everyone" He texted me two days later. "Delete everything I told you" Too late.

Lunar

1,652,836 Aufrufe • vor 2 Monaten

An Anthropic engineer left his laptop unlocked at a cafe in SF I wasn't trying to look. But the screen was right there. A Polymarket terminal. Live trades scrolling. Green numbers. And a model ID I'd never seen before - claude-opus-4-7. He came back. Saw me staring. Didn't close it. "You trade?" I told him I run 8 Claude agents on Polymarket. He sat down like I just said a password. "What model are you on?" 4.6. He shook his head. "We've been running 4.7 internally for two weeks. Your disposition filter - what is it?" 0.70. He laughed. "Ours is 0.82. Below that you're literally copying noise. But on 4.7 you can push it to 0.58. The model filters its own noise now" He turned his laptop. A spreadsheet. 47 wallets ranked by exit quality. "Not win rate. Exit quality. Top wallets capture 86% of the move and cut at 12%. Everyone else captures 58% and holds losers past 40%" I've never heard anyone break it down like that. He opened one repo. 86 million trades. Every wallet. Every entry. Every exit. Public. Then a second. "Three commands. Every order book. No API key. 4.7 scores them in seconds. Not minutes. Seconds" I asked why he's telling me this. "Because my side bot makes more than my Anthropic salary and I want the liquidity to stay deep" He showed me his phone. +$47,000. 38 days On a model that wasn't even public yet. I asked when 4.7 drops. He said "this week". It dropped today. I switched all 8 agents in 10 minutes. Rebuilt everything that night. First week: +$2,800 Second: +$2,400 Third: +$2,100 Fourth: +$2,100 +$9,400 in 28 days. 191 trades. 73% win rate. Avg hold 4h. Crypto +$3,740 Weather +$2,910 Politics +$1,850 Macro +$1,200 Sports -$300 Killed it after day 2 Capital velocity: 47x. Kelly f+ 0.068. Max drawdown -2.1%. Bot for those who don't build: I asked if I could post this. "Don't use my name and we're good" Done.
0:21

An Anthropic engineer left his laptop unlocked at a cafe in SF I wasn't trying to look. But the screen was right there. A Polymarket terminal. Live trades scrolling. Green numbers. And a model ID I'd never seen before - claude-opus-4-7. He came back. Saw me staring. Didn't close it. "You trade?" I told him I run 8 Claude agents on Polymarket. He sat down like I just said a password. "What model are you on?" 4.6. He shook his head. "We've been running 4.7 internally for two weeks. Your disposition filter - what is it?" 0.70. He laughed. "Ours is 0.82. Below that you're literally copying noise. But on 4.7 you can push it to 0.58. The model filters its own noise now" He turned his laptop. A spreadsheet. 47 wallets ranked by exit quality. "Not win rate. Exit quality. Top wallets capture 86% of the move and cut at 12%. Everyone else captures 58% and holds losers past 40%" I've never heard anyone break it down like that. He opened one repo. 86 million trades. Every wallet. Every entry. Every exit. Public. Then a second. "Three commands. Every order book. No API key. 4.7 scores them in seconds. Not minutes. Seconds" I asked why he's telling me this. "Because my side bot makes more than my Anthropic salary and I want the liquidity to stay deep" He showed me his phone. +$47,000. 38 days On a model that wasn't even public yet. I asked when 4.7 drops. He said "this week". It dropped today. I switched all 8 agents in 10 minutes. Rebuilt everything that night. First week: +$2,800 Second: +$2,400 Third: +$2,100 Fourth: +$2,100 +$9,400 in 28 days. 191 trades. 73% win rate. Avg hold 4h. Crypto +$3,740 Weather +$2,910 Politics +$1,850 Macro +$1,200 Sports -$300 Killed it after day 2 Capital velocity: 47x. Kelly f+ 0.068. Max drawdown -2.1%. Bot for those who don't build: I asked if I could post this. "Don't use my name and we're good" Done.

rari

71,528 Aufrufe • vor 2 Monaten

an ex-Jane Street trader told me a paycheck is the most overpriced contract there is we were at a rooftop bar in tel aviv. friend of a friend. i told him i work in tech. $4,200/month after taxes. he showed me his phone. a terminal. green numbers. +$47,000. last 30 days. "Claude and one GitHub repo. that's the whole setup" i asked what he means. "copy-trading. but not how everyone else does it" he explained it fast. like he'd said it a hundred times. "most people copy wallets by win rate. useless. a wallet with 85% wins can still lose money. what matters is disposition ratio" he pulled up a wallet. 91% win rate on crypto. 14% on politics. "copy everything - you bleed. copy only their crypto - top 1% of the platform" i went home. couldn't sleep. opened Claude at 2am. "build a copy engine. filter by disposition ratio above 0.70. copy only dominant category per wallet. kelly sizing" gave it this: almost 1k stars. execution engine. order book logic. position sizing. open source. Claude rebuilt the whole system in one night. first week: +$3,200. second: +$2,900. third: +$2,100. fourth: +$1,648. +$9,848 total. 313 trades. 80% win rate. avg hold 5h. the wallets it copies now aren't the highest win rate. they're the best exits. 0xe41f8...b has $43.4K profit. tech only. disposition 0.82. 0xa22c3...d has $40.2K. crypto only. disposition 0.79. 0xf88d2...c has $22.4K. macro only. disposition 0.74. category P&L: > crypto +$974 > weather +$749 > politics +$409 > macro +$353 > sports -$50. killed it capital velocity: 51x. kelly f+ 0.073. drawdown -2.9%. copytrade: saw him two weeks later. showed him the terminal. he scrolled through it. "you built this in one night?" Claude did. "my old desk had 14 people on this" mine costs less than a Netflix subscription. he nodded. "quit your job" i'm thinking about it.
0:33

an ex-Jane Street trader told me a paycheck is the most overpriced contract there is we were at a rooftop bar in tel aviv. friend of a friend. i told him i work in tech. $4,200/month after taxes. he showed me his phone. a terminal. green numbers. +$47,000. last 30 days. "Claude and one GitHub repo. that's the whole setup" i asked what he means. "copy-trading. but not how everyone else does it" he explained it fast. like he'd said it a hundred times. "most people copy wallets by win rate. useless. a wallet with 85% wins can still lose money. what matters is disposition ratio" he pulled up a wallet. 91% win rate on crypto. 14% on politics. "copy everything - you bleed. copy only their crypto - top 1% of the platform" i went home. couldn't sleep. opened Claude at 2am. "build a copy engine. filter by disposition ratio above 0.70. copy only dominant category per wallet. kelly sizing" gave it this: almost 1k stars. execution engine. order book logic. position sizing. open source. Claude rebuilt the whole system in one night. first week: +$3,200. second: +$2,900. third: +$2,100. fourth: +$1,648. +$9,848 total. 313 trades. 80% win rate. avg hold 5h. the wallets it copies now aren't the highest win rate. they're the best exits. 0xe41f8...b has $43.4K profit. tech only. disposition 0.82. 0xa22c3...d has $40.2K. crypto only. disposition 0.79. 0xf88d2...c has $22.4K. macro only. disposition 0.74. category P&L: > crypto +$974 > weather +$749 > politics +$409 > macro +$353 > sports -$50. killed it capital velocity: 51x. kelly f+ 0.073. drawdown -2.9%. copytrade: saw him two weeks later. showed him the terminal. he scrolled through it. "you built this in one night?" Claude did. "my old desk had 14 people on this" mine costs less than a Netflix subscription. he nodded. "quit your job" i'm thinking about it.

Hanako

310,636 Aufrufe • vor 2 Monaten

a drunk Jane Street quant drew his trading system on a napkin it started as a joke. someone's friend at a bar in lower manhattan. he mentioned prediction markets and this guy lit up. "i do this for a living. for a firm you've heard of" two rounds later he was drawing on a napkin. boxes. arrows. labels. "the crowd repeats six mistakes every single day. we just built detectors for each one" > base rate neglect > mispriced conditionals > thin book slippage > platform divergence > favourite bias > time decay "that's the whole system" i asked what it costs to run. "200 quants. proprietary infra. about $2M/year just to keep it on" i laughed. photographed the napkin. went home. 3am. couldn't sleep. opened Claude. fed it the photo and one GitHub repo - poly_data. 86 million trades. every wallet. every entry. said: build detectors for all six. by 4AM my laptop fan was screaming. 400+ markets scanning per hour. 8 agents deployed. terminal live. i didn't understand half of what was on screen. but the P&L number kept climbing. +$312 by breakfast. +$1,900 by end of day one. +$7,400 by end of week one. +$16,474 right now. 383 trades. 81% win rate. the scanner finds gaps like: > AAPL earnings - crowd says 77%. data says 91%. > Gold $2600 - crowd says 44%. model says 60%. > Senate filibuster - crowd says 68%. base rate says 82%. enters. waits 6 hours average. exits when the gap closes. politics at 93%. weather 79%. crypto 72%. eight agents split the work. one mirrors whales. one catches Binance lag. one fades favourites. one bleeds theta on dying contracts. his firm runs this with 200 people. mine runs on $25/month. copytrade here: texted him the screenshot yesterday. long pause. "you're not supposed to be able to do this" i know.
0:27

a drunk Jane Street quant drew his trading system on a napkin it started as a joke. someone's friend at a bar in lower manhattan. he mentioned prediction markets and this guy lit up. "i do this for a living. for a firm you've heard of" two rounds later he was drawing on a napkin. boxes. arrows. labels. "the crowd repeats six mistakes every single day. we just built detectors for each one" > base rate neglect > mispriced conditionals > thin book slippage > platform divergence > favourite bias > time decay "that's the whole system" i asked what it costs to run. "200 quants. proprietary infra. about $2M/year just to keep it on" i laughed. photographed the napkin. went home. 3am. couldn't sleep. opened Claude. fed it the photo and one GitHub repo - poly_data. 86 million trades. every wallet. every entry. said: build detectors for all six. by 4AM my laptop fan was screaming. 400+ markets scanning per hour. 8 agents deployed. terminal live. i didn't understand half of what was on screen. but the P&L number kept climbing. +$312 by breakfast. +$1,900 by end of day one. +$7,400 by end of week one. +$16,474 right now. 383 trades. 81% win rate. the scanner finds gaps like: > AAPL earnings - crowd says 77%. data says 91%. > Gold $2600 - crowd says 44%. model says 60%. > Senate filibuster - crowd says 68%. base rate says 82%. enters. waits 6 hours average. exits when the gap closes. politics at 93%. weather 79%. crypto 72%. eight agents split the work. one mirrors whales. one catches Binance lag. one fades favourites. one bleeds theta on dying contracts. his firm runs this with 200 people. mine runs on $25/month. copytrade here: texted him the screenshot yesterday. long pause. "you're not supposed to be able to do this" i know.

Hanako

244,194 Aufrufe • vor 2 Monaten

An Anthropic safety researcher closed her laptop when she saw my screen at Philz Coffee I was running my Polymarket bot from the corner table. She was in line. Looked over my shoulder. Stopped moving. "That's not a normal trading app. What model is that running on" I told her. Claude Code. Four repos. $25 a month. She sat down without asking. "I work on the alignment team. We test Claude for exactly this kind of autonomous behavior. You're letting it find its own trading signals" Not just signals. Wallets. 86 million trades. Every wallet. Every entry. Every exit. "You're feeding Claude raw wallet data and letting it identify which traders consistently win. Then cloning their behavior" She said it slowly. Like she was writing an internal report in her head. Claude Code finds the top wallets. Reverse-engineers their timing. Copies their entries. Then exits before they do. "Before they do?" My bot cuts at 85% of expected move or on a 3x volume spike. Top wallets exit before resolution 91% of the time. They capture 86% of the move. Losers hold to 58%. She put her coffee down. "How did you get Claude to learn exit timing on its own" I showed her the second repo. Three commands. 500+ markets. No API key. Claude scores them in 20 minutes. "We have 14 people stress-testing Claude's autonomous capabilities. You're just using them" My setup: Claude API - $20/mo VPS - $5/mo poly_data - free polymarket-cli - free 19 days. 4 agents. 74% win rate. Copytrade here: She stared at the screen for a long time. "This is literally what our red team simulates. Except you actually deployed it" She emailed me two days later. "Our policy team found your post. Please take it down" Too late.
0:20

An Anthropic safety researcher closed her laptop when she saw my screen at Philz Coffee I was running my Polymarket bot from the corner table. She was in line. Looked over my shoulder. Stopped moving. "That's not a normal trading app. What model is that running on" I told her. Claude Code. Four repos. $25 a month. She sat down without asking. "I work on the alignment team. We test Claude for exactly this kind of autonomous behavior. You're letting it find its own trading signals" Not just signals. Wallets. 86 million trades. Every wallet. Every entry. Every exit. "You're feeding Claude raw wallet data and letting it identify which traders consistently win. Then cloning their behavior" She said it slowly. Like she was writing an internal report in her head. Claude Code finds the top wallets. Reverse-engineers their timing. Copies their entries. Then exits before they do. "Before they do?" My bot cuts at 85% of expected move or on a 3x volume spike. Top wallets exit before resolution 91% of the time. They capture 86% of the move. Losers hold to 58%. She put her coffee down. "How did you get Claude to learn exit timing on its own" I showed her the second repo. Three commands. 500+ markets. No API key. Claude scores them in 20 minutes. "We have 14 people stress-testing Claude's autonomous capabilities. You're just using them" My setup: Claude API - $20/mo VPS - $5/mo poly_data - free polymarket-cli - free 19 days. 4 agents. 74% win rate. Copytrade here: She stared at the screen for a long time. "This is literally what our red team simulates. Except you actually deployed it" She emailed me two days later. "Our policy team found your post. Please take it down" Too late.

Lunar

231,073 Aufrufe • vor 2 Monaten

An ex-OpenAI engineer walked up to me at a meetup in SF I was showing my Polymarket terminal. He looked at the screen for ten seconds and said one sentence. "You're trading blind. The data is sitting in the open and you're writing prompts" I didn't understand. What data. He took my laptop. Opened one repository. 86 million trades. Every wallet. Every entry. Every exit. The entire Polymarket history since day one. "At OpenAI models don't guess. They read. Connect Claude Code to this dataset and say - find every wallet with a win rate above 70% and more than 100 trades" I asked - why Claude and not GPT? He looked at me like I was an idiot. "Because Claude Code connects to the repo directly. It reads the entire codebase. It's not a chatbox. It's a runtime" That evening I connected it. Claude pulled 47 wallets in 4 minutes. Average profit: $214K. Hold time: 7 hours. 91% close their position BEFORE resolution. Top wallets capture 86% of the move and cut losers at 12%. Everyone else - 58% of profit and hold losers to 41%. Same exact entries. The exits make it a completely different game. "Now connect the scanner" Three commands - the bot sees 500+ markets in real time. No API key. Read-only. Claude built the scoring in 20 minutes: Gap between price and model > 7 cents. Book depth > $500. Resolution in 4-48 hours. 93% of markets get killed instantly. Only the fat ones survive. "Now this is the important part" - he sent me an article where a guy built a full bot over a weekend from these same repos Copytrade here: Three exit triggers: Target 85% of expected move. Volume spike x3 - smart money leaving. 24 hours of silence - thesis is dead. I copied the whole stack. VPS $5. Claude $20. Total $25 a month. No team. No office. No Bloomberg. 16 days. 187 trades. 71% win rate. $800 seed. +$8,700. I sent him my screen. He replied a day later. "You just replicated for $25 what cost us six months and 11 people" I said - thanks for the tip. "Delete this chat" Too late.
0:24

An ex-OpenAI engineer walked up to me at a meetup in SF I was showing my Polymarket terminal. He looked at the screen for ten seconds and said one sentence. "You're trading blind. The data is sitting in the open and you're writing prompts" I didn't understand. What data. He took my laptop. Opened one repository. 86 million trades. Every wallet. Every entry. Every exit. The entire Polymarket history since day one. "At OpenAI models don't guess. They read. Connect Claude Code to this dataset and say - find every wallet with a win rate above 70% and more than 100 trades" I asked - why Claude and not GPT? He looked at me like I was an idiot. "Because Claude Code connects to the repo directly. It reads the entire codebase. It's not a chatbox. It's a runtime" That evening I connected it. Claude pulled 47 wallets in 4 minutes. Average profit: $214K. Hold time: 7 hours. 91% close their position BEFORE resolution. Top wallets capture 86% of the move and cut losers at 12%. Everyone else - 58% of profit and hold losers to 41%. Same exact entries. The exits make it a completely different game. "Now connect the scanner" Three commands - the bot sees 500+ markets in real time. No API key. Read-only. Claude built the scoring in 20 minutes: Gap between price and model > 7 cents. Book depth > $500. Resolution in 4-48 hours. 93% of markets get killed instantly. Only the fat ones survive. "Now this is the important part" - he sent me an article where a guy built a full bot over a weekend from these same repos Copytrade here: Three exit triggers: Target 85% of expected move. Volume spike x3 - smart money leaving. 24 hours of silence - thesis is dead. I copied the whole stack. VPS $5. Claude $20. Total $25 a month. No team. No office. No Bloomberg. 16 days. 187 trades. 71% win rate. $800 seed. +$8,700. I sent him my screen. He replied a day later. "You just replicated for $25 what cost us six months and 11 people" I said - thanks for the tip. "Delete this chat" Too late.

Lunar

480,397 Aufrufe • vor 2 Monaten

US export-control disable Claude Fable 5 overnight. Not for what it said - for what one trader did with it. He turned an $11,000 wreck into $300,000 profit in five days. Then they pulled Claude Fable 5 from public hands. Too late - his bot's still running. His wallet: Every other Claude model still works. The two best ones don't. Anthropic's calling it a misunderstanding. Maybe. But think about what gets a model killed by export control overnight, no warning. Here's what Fable 5 actually did for him. It doesn't predict price. It computes the whole probability surface of a market - every outcome, weighted, faster than the market can correct itself. It finds the bets the crowd writes off as impossible, runs the real math, and buys them for pennies when the math disagrees. One of those just paid x205. A regular guy, a model this sharp, a market no regulator polices. They closed that door overnight - one night too late. He's holding one of the last builds that ever touched a live market. It's still on Polymarket right now, hunting the same mispriced tails. Fable 5 is locked. The algorithm it wrote is loose. You can't open the model anymore. You can still copy the wallet it built in two clicks:
0:46

US export-control disable Claude Fable 5 overnight. Not for what it said - for what one trader did with it. He turned an $11,000 wreck into $300,000 profit in five days. Then they pulled Claude Fable 5 from public hands. Too late - his bot's still running. His wallet: Every other Claude model still works. The two best ones don't. Anthropic's calling it a misunderstanding. Maybe. But think about what gets a model killed by export control overnight, no warning. Here's what Fable 5 actually did for him. It doesn't predict price. It computes the whole probability surface of a market - every outcome, weighted, faster than the market can correct itself. It finds the bets the crowd writes off as impossible, runs the real math, and buys them for pennies when the math disagrees. One of those just paid x205. A regular guy, a model this sharp, a market no regulator polices. They closed that door overnight - one night too late. He's holding one of the last builds that ever touched a live market. It's still on Polymarket right now, hunting the same mispriced tails. Fable 5 is locked. The algorithm it wrote is loose. You can't open the model anymore. You can still copy the wallet it built in two clicks:

cvxv666

77,026 Aufrufe • vor 22 Tagen

Ran into an Anthropic engineer at an airport lounge in Frankfurt last month. SFO delayed six hours. Nothing to do. I had my laptop open. Bot running. Three agents scoring markets in the background. Guy sits two seats down. Glances at my screen. Looks away. Looks back. "Is that Polymarket?" I said yeah. "You're running it with Claude Code?" I nodded. He pulled his chair closer. "I'm on the agent team. We stress-test this exact setup internally. You built it in your kitchen" I asked what stress-test meant. "Red team. We simulate what happens when someone gives Claude raw wallet data and tells it to find the winners. You're running the thing we were scared about" I showed him the scanner. One prompt. Find every wallet with 100 plus trades, 70 plus win rate. Rank by profit. Export top 50. Claude chewed through 14,000 wallets in under 4 minutes. Came back with 47. Top 20 made more than the bottom 13,000 combined. "That's not a distribution. That's a hit list" I said yeah. "And the scoring function?" Claude wrote it. I just wrapped it in an if-statement. A fill landed on screen mid-conversation. +$128 on an ETH dominance market. He stared at it. "How does it decide to enter" Three agents. Shared wallet. No shared memory. Arbitrage, convergence, whale copy. 2 of 3 agree, full size. 1 alone, half. Disagree, no trade. That consensus filter alone killed 40% of losing trades. "And the exit" The 47 whales never hold to settlement. 91% exit early. Capture 73% of max. I cut at 85% of the expected move. Or a 3x volume spike. Whichever comes first. "So you built a whale copy bot that exits before the whales" Yeah. He put his coffee down. "That's exactly the thing we gamed out in March. Management killed the internal version on policy. Nobody outside the team should have shipped this yet" I asked what he was going to do. "Fly home and pretend I didn't see it" $600 seed. 31 days. Net +$35,200. 479 trades. 75% win rate. Sharpe 2.63. I haven't touched the bot in three weeks. Copytrade for those who don't want to build - Before boarding he slid a card across the table. "When the next agent model drops, rerun your prompt. You'll see what I mean"
0:27

Ran into an Anthropic engineer at an airport lounge in Frankfurt last month. SFO delayed six hours. Nothing to do. I had my laptop open. Bot running. Three agents scoring markets in the background. Guy sits two seats down. Glances at my screen. Looks away. Looks back. "Is that Polymarket?" I said yeah. "You're running it with Claude Code?" I nodded. He pulled his chair closer. "I'm on the agent team. We stress-test this exact setup internally. You built it in your kitchen" I asked what stress-test meant. "Red team. We simulate what happens when someone gives Claude raw wallet data and tells it to find the winners. You're running the thing we were scared about" I showed him the scanner. One prompt. Find every wallet with 100 plus trades, 70 plus win rate. Rank by profit. Export top 50. Claude chewed through 14,000 wallets in under 4 minutes. Came back with 47. Top 20 made more than the bottom 13,000 combined. "That's not a distribution. That's a hit list" I said yeah. "And the scoring function?" Claude wrote it. I just wrapped it in an if-statement. A fill landed on screen mid-conversation. +$128 on an ETH dominance market. He stared at it. "How does it decide to enter" Three agents. Shared wallet. No shared memory. Arbitrage, convergence, whale copy. 2 of 3 agree, full size. 1 alone, half. Disagree, no trade. That consensus filter alone killed 40% of losing trades. "And the exit" The 47 whales never hold to settlement. 91% exit early. Capture 73% of max. I cut at 85% of the expected move. Or a 3x volume spike. Whichever comes first. "So you built a whale copy bot that exits before the whales" Yeah. He put his coffee down. "That's exactly the thing we gamed out in March. Management killed the internal version on policy. Nobody outside the team should have shipped this yet" I asked what he was going to do. "Fly home and pretend I didn't see it" $600 seed. 31 days. Net +$35,200. 479 trades. 75% win rate. Sharpe 2.63. I haven't touched the bot in three weeks. Copytrade for those who don't want to build - Before boarding he slid a card across the table. "When the next agent model drops, rerun your prompt. You'll see what I mean"

rari

618,409 Aufrufe • vor 2 Monaten

Private code leaked on GitHub for 6 hours. Someone pushed to public by mistake. I cloned it before the dev woke up and deleted everything. After adding it to my C++ scanner, first catch was a wallet printing $19.5K per day. CRYINGLITTLEBABY. $379K profit in a single month. 100% win rate. → Wallet: I scrolled the transaction history for ten minutes looking for red. One mistake. One bad trade. Something to prove the data was broken. Nothing. Solid wall of green. My terminal cross referenced the timestamps with spot data. That is when I saw the pattern. This bot does not trade. It collects. BTC drops 0.3% on Binance. The move already happened. Fact. But Polymarket still shows old odds. The DOWN outcome sits mispriced at 30 cents. Bot slams the order. Waits. Window closes. Collects $1. Paying 30 cents for a guaranteed dollar. Not a prediction. A receipt for something that already occurred. Before the GitHub upgrade my scanner would have missed this. Too many wallets. Too much noise. The new code filters by execution timing. Catches only the ones exploiting the lag. CRYINGLITTLEBABY runs this pattern all day. $19.5K average. No losing days in the entire history. Spent six months building my terminal from scratch. One leaked repo turned it into something that actually finds money. The best code is not always the code you write. Sometimes it is the code you grab at 3am before someone deletes it forever.
0:51

Private code leaked on GitHub for 6 hours. Someone pushed to public by mistake. I cloned it before the dev woke up and deleted everything. After adding it to my C++ scanner, first catch was a wallet printing $19.5K per day. CRYINGLITTLEBABY. $379K profit in a single month. 100% win rate. → Wallet: I scrolled the transaction history for ten minutes looking for red. One mistake. One bad trade. Something to prove the data was broken. Nothing. Solid wall of green. My terminal cross referenced the timestamps with spot data. That is when I saw the pattern. This bot does not trade. It collects. BTC drops 0.3% on Binance. The move already happened. Fact. But Polymarket still shows old odds. The DOWN outcome sits mispriced at 30 cents. Bot slams the order. Waits. Window closes. Collects $1. Paying 30 cents for a guaranteed dollar. Not a prediction. A receipt for something that already occurred. Before the GitHub upgrade my scanner would have missed this. Too many wallets. Too much noise. The new code filters by execution timing. Catches only the ones exploiting the lag. CRYINGLITTLEBABY runs this pattern all day. $19.5K average. No losing days in the entire history. Spent six months building my terminal from scratch. One leaked repo turned it into something that actually finds money. The best code is not always the code you write. Sometimes it is the code you grab at 3am before someone deletes it forever.

Marlow

669,974 Aufrufe • vor 5 Monaten

A Google DeepMind researcher cornered me at a bar in Hayes Valley I was showing my Polymarket PNL to a friend. She leaned over. Didn't introduce herself. "That's not a trading app. Show me your stack" I told her. Claude Code. Four repos. $25 a month. She set down her drink. "We tested this internally. You connect Claude directly to a dataset. It builds its own detectors. But nobody ships it because compliance kills everything" I asked what she meant. She took my phone. Opened one link. 86 million trades. Every wallet. Every entry. Every exit. "You don't tell Claude what to look for. It finds the wallets that win. Then it finds WHY they win. Then it copies the pattern" Her team spent 9 months building this for a hedge fund. 14 people. $2M budget. "The part that took us the longest - exit logic. Everyone thinks entries matter. They don't. Exits are the entire game" I told her my bot cuts at 85% of expected move or on a 3x volume spike. She went quiet. "Who taught you that" Claude Code found it in poly_data. Top wallets exit before resolution 91% of the time. They capture the move and leave. She opened another link. "This is the scanner. Three commands. 500+ markets. No API key. Claude scores them in 20 minutes" "That's our exact infra. Except it took us 9 months and you did it in a weekend" My setup: Claude API - $20/mo VPS - $5/mo poly_data - free polymarket-cli - free 19 days. 4 agents. 74% win rate. Copytrade here: I showed her the article where I broke down every repo, every command, every dollar. She read it for five minutes. Then: "You just open-sourced our entire pipeline" She texted me the next day. "My team lead saw your thread. Take it down" Too late.
0:21

A Google DeepMind researcher cornered me at a bar in Hayes Valley I was showing my Polymarket PNL to a friend. She leaned over. Didn't introduce herself. "That's not a trading app. Show me your stack" I told her. Claude Code. Four repos. $25 a month. She set down her drink. "We tested this internally. You connect Claude directly to a dataset. It builds its own detectors. But nobody ships it because compliance kills everything" I asked what she meant. She took my phone. Opened one link. 86 million trades. Every wallet. Every entry. Every exit. "You don't tell Claude what to look for. It finds the wallets that win. Then it finds WHY they win. Then it copies the pattern" Her team spent 9 months building this for a hedge fund. 14 people. $2M budget. "The part that took us the longest - exit logic. Everyone thinks entries matter. They don't. Exits are the entire game" I told her my bot cuts at 85% of expected move or on a 3x volume spike. She went quiet. "Who taught you that" Claude Code found it in poly_data. Top wallets exit before resolution 91% of the time. They capture the move and leave. She opened another link. "This is the scanner. Three commands. 500+ markets. No API key. Claude scores them in 20 minutes" "That's our exact infra. Except it took us 9 months and you did it in a weekend" My setup: Claude API - $20/mo VPS - $5/mo poly_data - free polymarket-cli - free 19 days. 4 agents. 74% win rate. Copytrade here: I showed her the article where I broke down every repo, every command, every dollar. She read it for five minutes. Then: "You just open-sourced our entire pipeline" She texted me the next day. "My team lead saw your thread. Take it down" Too late.

Lunar

124,203 Aufrufe • vor 2 Monaten

I asked Claude: I have $500. Where do I put it to make the most money possible? Expected the usual. Stocks. ETFs. Maybe real estate. Instead Claude said: Go to GitHub. Search polymarket-1-cent-bot. Find the repo with the most stars. Deploy it with your $500. Most positions will expire at zero. The ones that hit pay 80x. Someone is already running this at scale. Search distinct-baguette. I searched GitHub first. Found a 340 star repo. Last commit 3 days ago. 47 lines of Python. README had one line: buys chaos at 1 cent, sells certainty at $1. Then I searched the wallet. distinct-baguette. $805,888 profit. 45,739 trades. Joined October 2025. Bio: Hi Isola. I asked Claude: how does the math work? $500 across 50 positions at 1 cent each. 47 expire worthless. You lose $470. Three hit. Each pays $100. You walk away with $300 net profit. If four hit you double. Five and you're distinct-baguette. I put in $200. Copied the repo. Went to sleep. One hit. $200 → $1,340. I asked Claude: why didn't you just say buy an index fund? You asked for the most money possible. Not the most comfortable answer. GitHub repo still public. Wallet still running. Bio still says Hi Isola.
0:57

I asked Claude: I have $500. Where do I put it to make the most money possible? Expected the usual. Stocks. ETFs. Maybe real estate. Instead Claude said: Go to GitHub. Search polymarket-1-cent-bot. Find the repo with the most stars. Deploy it with your $500. Most positions will expire at zero. The ones that hit pay 80x. Someone is already running this at scale. Search distinct-baguette. I searched GitHub first. Found a 340 star repo. Last commit 3 days ago. 47 lines of Python. README had one line: buys chaos at 1 cent, sells certainty at $1. Then I searched the wallet. distinct-baguette. $805,888 profit. 45,739 trades. Joined October 2025. Bio: Hi Isola. I asked Claude: how does the math work? $500 across 50 positions at 1 cent each. 47 expire worthless. You lose $470. Three hit. Each pays $100. You walk away with $300 net profit. If four hit you double. Five and you're distinct-baguette. I put in $200. Copied the repo. Went to sleep. One hit. $200 → $1,340. I asked Claude: why didn't you just say buy an index fund? You asked for the most money possible. Not the most comfortable answer. GitHub repo still public. Wallet still running. Bio still says Hi Isola.

Marlow

294,364 Aufrufe • vor 3 Monaten

Sam Altman on GPT 5: "This morning I was testing our new model and I got a question. I got emailed a question that I didn't quite understand. And I put it in the model, this is GPT-5, and it answered it perfectly. And I really kind of sat back in my chair and I was just like, oh man, here it is moment... I felt like useless relative to the AI in this thing that I felt like I should have been able to do and I couldn't. It was really hard. But the AI just did it like that. It was a weird feeling."
0:46

Sam Altman on GPT 5: "This morning I was testing our new model and I got a question. I got emailed a question that I didn't quite understand. And I put it in the model, this is GPT-5, and it answered it perfectly. And I really kind of sat back in my chair and I was just like, oh man, here it is moment... I felt like useless relative to the AI in this thing that I felt like I should have been able to do and I couldn't. It was really hard. But the AI just did it like that. It was a weird feeling."

Chris

5,664,499 Aufrufe • vor 11 Monaten

Karpathy said something you'll regret ignoring: "We have to keep the AI on the leash. I'm still the bottleneck. I have to make sure this thing isn't introducing bugs and that there's no security issues." He said it at YC talk last year, when the worry was reliability. The models hallucinated and made mistakes no human would, so the leash implied keeping yourself in the loop and checking the output before trusting it. The models are far better now, and the line still holds, for a reason he was not focused on back then. Even a model that writes flawless code today still has no idea who is allowed to run it. Correctness and authorization are different problems, and only correctness improves as the model improves. A perfect agent still hands a tool where anyone can do anything, because permission was never part of the task. I actually tested this in practice with Claude Code. I asked it to build a small internal tool with a button that issues account credits. It worked first try, and running it locally, the credit applied the instant I clicked. Nothing decided who was allowed to click it. The agent wrote the right logic and displayed a success notification. It never checked whether the caller had the right, whether it should pause for a human, or whether anything was logged. And this is not a bug a smarter model can outgrow because the leash was never in the code. Identity, permissions, and audit live in the system that runs the app, not in what the agent generates. To solve this, I took the exact same bundle and hosted it on Retool. The credit write that fired silently on my laptop now stopped at an approval gate, resolved to a real identity through SSO, and landed in an audit log. I wrote none of it. The app inherited the entire boundary the moment it was deployed, and the video shows the before and after. You can try it yourself here: I also wrote a detailed breakdown of the whole thing in my recent article, and I worked with the team to put this together. It walks through the build, the exact moment the credit write went through on my laptop with nobody checking, and then what changed when the same app ran on Retool. It also covers why this is a property of the runtime and not something a better model fixes, which is why devs typically miss this. The article is quoted below.
0:38

Karpathy said something you'll regret ignoring: "We have to keep the AI on the leash. I'm still the bottleneck. I have to make sure this thing isn't introducing bugs and that there's no security issues." He said it at YC talk last year, when the worry was reliability. The models hallucinated and made mistakes no human would, so the leash implied keeping yourself in the loop and checking the output before trusting it. The models are far better now, and the line still holds, for a reason he was not focused on back then. Even a model that writes flawless code today still has no idea who is allowed to run it. Correctness and authorization are different problems, and only correctness improves as the model improves. A perfect agent still hands a tool where anyone can do anything, because permission was never part of the task. I actually tested this in practice with Claude Code. I asked it to build a small internal tool with a button that issues account credits. It worked first try, and running it locally, the credit applied the instant I clicked. Nothing decided who was allowed to click it. The agent wrote the right logic and displayed a success notification. It never checked whether the caller had the right, whether it should pause for a human, or whether anything was logged. And this is not a bug a smarter model can outgrow because the leash was never in the code. Identity, permissions, and audit live in the system that runs the app, not in what the agent generates. To solve this, I took the exact same bundle and hosted it on Retool. The credit write that fired silently on my laptop now stopped at an approval gate, resolved to a real identity through SSO, and landed in an audit log. I wrote none of it. The app inherited the entire boundary the moment it was deployed, and the video shows the before and after. You can try it yourself here: I also wrote a detailed breakdown of the whole thing in my recent article, and I worked with the team to put this together. It walks through the build, the exact moment the credit write went through on my laptop with nobody checking, and then what changed when the same app ran on Retool. It also covers why this is a property of the runtime and not something a better model fixes, which is why devs typically miss this. The article is quoted below.

Akshay 🚀

42,511 Aufrufe • vor 13 Tagen

I gave Claude $20,000 to LIVE trade the crypto markets. I was not expecting what happened next... It printed $8,000+ in its first week of live trading. The best part is, literally anyone can copy this setup. In the comments below, I just posted the full architecture (free).
1:10

I gave Claude $20,000 to LIVE trade the crypto markets. I was not expecting what happened next... It printed $8,000+ in its first week of live trading. The best part is, literally anyone can copy this setup. In the comments below, I just posted the full architecture (free).

Miles Deutscher

70,489 Aufrufe • vor 1 Monat

a drunk Chinese friend who works as a quant slipped me a napkin with 8 words on it we were at a rooftop in tel aviv. he was three cocktails past the professional filter. i mentioned prediction markets. he laughed. grabbed a napkin from the bar. scribbled in english. "disposition. category. velocity. cross-market. news. volume. whale. theta." i asked what it means. "eight signals. every mistake on these markets is one of them. we run detectors in parallel. enter only when three or more fire" i asked how long it took to build. "two years. fourteen quants" i asked if i could just do it myself. he laughed harder. "with Claude? tonight probably" i don't think he meant to say that part. he looked at the napkin like he regretted writing it. too late. i pocketed it. went home. opened Claude at 11PM. typed the eight words. "build a parallel detector grid on prediction markets. 8 signals. enter only when 3 or more fire simultaneously" Claude needed a backbone. i gave it: 610 stars. market making bot. order book execution on both sides. position sizing. the foundation for anything that needs to place real orders fast. terminal was live by 2AM. the radar lights up 14 times per hour. most signals fire alone. rare ones stack. last week: > BRICS summit - 74c. 4 signals. +24c in 3h. > Artemis launch - 63c. 5 signals. +33c in 5h10m. > Helium-3 demo - 68c. 3 signals. +14c in 48m. > Waymo public - 67c. 4 signals. +8c in 6m. sentiment feed reads 4 platforms: - X/CT: 56% bearish - Reddit: 53% - Telegram: 17% - Discord: 75% when sentiment diverges from flow pressure by 20%+ the contrarian signal fires. 83% win rate on those. cross-market delta catches gaps: BTC 92K vs Kalshi 7.7c Fed cut vs PredictIt 7.2c AAPL>180 vs Manifold 5.7c oracle delta tracks NOAA, AP, CoinGecko, Reuters. when a feed moves before the market adjusts, the bot enters on the lag. 6 wallets copied by dominant category: > 0xe41f8...b2 - $483 profit. 83% wr. > 0xa22c3...d9 - $890. 67%. > voltwhale - $593. 55%. > 0xf88d2...c4 - $830. 73%. 412 trades. 76% win rate. avg hold 5h22m. best $232. worst -$37. 9 consecutive wins. +$12,640. from a $900 seed. 4 weeks live. copytrade setup: i haven't seen him since that night. the napkin is still in my drawer. i don't think he remembers writing it.
0:31

a drunk Chinese friend who works as a quant slipped me a napkin with 8 words on it we were at a rooftop in tel aviv. he was three cocktails past the professional filter. i mentioned prediction markets. he laughed. grabbed a napkin from the bar. scribbled in english. "disposition. category. velocity. cross-market. news. volume. whale. theta." i asked what it means. "eight signals. every mistake on these markets is one of them. we run detectors in parallel. enter only when three or more fire" i asked how long it took to build. "two years. fourteen quants" i asked if i could just do it myself. he laughed harder. "with Claude? tonight probably" i don't think he meant to say that part. he looked at the napkin like he regretted writing it. too late. i pocketed it. went home. opened Claude at 11PM. typed the eight words. "build a parallel detector grid on prediction markets. 8 signals. enter only when 3 or more fire simultaneously" Claude needed a backbone. i gave it: 610 stars. market making bot. order book execution on both sides. position sizing. the foundation for anything that needs to place real orders fast. terminal was live by 2AM. the radar lights up 14 times per hour. most signals fire alone. rare ones stack. last week: > BRICS summit - 74c. 4 signals. +24c in 3h. > Artemis launch - 63c. 5 signals. +33c in 5h10m. > Helium-3 demo - 68c. 3 signals. +14c in 48m. > Waymo public - 67c. 4 signals. +8c in 6m. sentiment feed reads 4 platforms: - X/CT: 56% bearish - Reddit: 53% - Telegram: 17% - Discord: 75% when sentiment diverges from flow pressure by 20%+ the contrarian signal fires. 83% win rate on those. cross-market delta catches gaps: BTC 92K vs Kalshi 7.7c Fed cut vs PredictIt 7.2c AAPL>180 vs Manifold 5.7c oracle delta tracks NOAA, AP, CoinGecko, Reuters. when a feed moves before the market adjusts, the bot enters on the lag. 6 wallets copied by dominant category: > 0xe41f8...b2 - $483 profit. 83% wr. > 0xa22c3...d9 - $890. 67%. > voltwhale - $593. 55%. > 0xf88d2...c4 - $830. 73%. 412 trades. 76% win rate. avg hold 5h22m. best $232. worst -$37. 9 consecutive wins. +$12,640. from a $900 seed. 4 weeks live. copytrade setup: i haven't seen him since that night. the napkin is still in my drawer. i don't think he remembers writing it.

Hanako

156,701 Aufrufe • vor 2 Monaten

I ran a quick test of Gemini Robotics ER 1.5 controlling my LeKiwi using the MCP I developed before. No extra code required - just specified a new model. Unfortunately, it doesn't work well, even though I had expected it to be the best, as it is the first major VLM model trained specifically for robotics applications. The video below is a relatively good attempt. It does understand what is going on and generates a reasonable action plan, but it is not doing well with the low-level control. I guess it was not trained for this and was used together with the VLA, so it is trained to generate more high-level commands like "grab this object and put it there" rather than doing low-level control specifying every robot movement step by step. Did anybody try to use it in a similar setup? What is your experience?
1:11

I ran a quick test of Gemini Robotics ER 1.5 controlling my LeKiwi using the MCP I developed before. No extra code required - just specified a new model. Unfortunately, it doesn't work well, even though I had expected it to be the best, as it is the first major VLM model trained specifically for robotics applications. The video below is a relatively good attempt. It does understand what is going on and generates a reasonable action plan, but it is not doing well with the low-level control. I guess it was not trained for this and was used together with the VLA, so it is trained to generate more high-level commands like "grab this object and put it there" rather than doing low-level control specifying every robot movement step by step. Did anybody try to use it in a similar setup? What is your experience?

Ilia

15,601 Aufrufe • vor 9 Monaten