Ricker's banner

Ricker

@0xRicker • 9,220 subscribers

Researcher & Contributor • Creator Partnerships & collabs welcome • DMs always open

Shorts

This trader has won 148 predictions in a row on Polymarket He only bets on No and always win Here's the cold math behind his $275k profit 148 closed positions. Every single one: Won His secret? He doesn't predict the future The math that makes this work: EV = P(No) × Payout − Cost When the market says 95¢, it implies 95% confidence You should try copy-trade him even with $10 using Kreo: He just identifies what the market is overestimating His strategy: • Find extreme price targets the crowd thinks are possible • Buy "No" when the odds reach 88-98¢ • Collect the premium like an insurance company He doesn't care about the asset He cares about the mispricing

This trader has won 148 predictions in a row on Polymarket He only bets on No and always win Here's the cold math behind his $275k profit 148 closed positions. Every single one: Won His secret? He doesn't predict the future The math that makes this work: EV = P(No) × Payout − Cost When the market says 95¢, it implies 95% confidence You should try copy-trade him even with $10 using Kreo: He just identifies what the market is overestimating His strategy: • Find extreme price targets the crowd thinks are possible • Buy "No" when the odds reach 88-98¢ • Collect the premium like an insurance company He doesn't care about the asset He cares about the mispricing

198,737 views

How to Train a Trading Bot on Real Indicators to $1000/day print Most losing bots blow up not from bad entries, but from trading when they shouldn't The formula: > RSI (context) + MACD (momentum) + Stochastic (timing) + EMA (structure) + OBV (confirmation) + VWAP (value) + Volatility (filter) = a bot that stops guessing and executes a framework 7 Indicators That Actually Work Also the fastest way to copy-trade anyone: 1. RSI - crowd emotion filter - RSI > 70 → market overheated → look for NO - RSI 80 = overheated, < 20 = oversold, %K crossing %D = entry trigger 4. EMA - price structure - Fast EMA (9) + slow EMA (21/50) - Price chopping around EMAs → stay out, price clearly trending → engage 5. OBV - real money flow - Price ↑ + OBV ↑ = strong trend - Price ↑ + OBV flat/↓ = weak move = trap 6. VWAP - fair market price - Price above VWAP = bullish bias, below = bearish bias - Key edge: deviation from VWAP → often reverts = high-probability entry setup 7. Volatility Filter - when NOT to trade - Uses ATR (Average True Range) - Rule: only trade within the optimal volatility range

How to Train a Trading Bot on Real Indicators to $1000/day print Most losing bots blow up not from bad entries, but from trading when they shouldn't The formula: > RSI (context) + MACD (momentum) + Stochastic (timing) + EMA (structure) + OBV (confirmation) + VWAP (value) + Volatility (filter) = a bot that stops guessing and executes a framework 7 Indicators That Actually Work Also the fastest way to copy-trade anyone: 1. RSI - crowd emotion filter - RSI > 70 → market overheated → look for NO - RSI 80 = overheated, < 20 = oversold, %K crossing %D = entry trigger 4. EMA - price structure - Fast EMA (9) + slow EMA (21/50) - Price chopping around EMAs → stay out, price clearly trending → engage 5. OBV - real money flow - Price ↑ + OBV ↑ = strong trend - Price ↑ + OBV flat/↓ = weak move = trap 6. VWAP - fair market price - Price above VWAP = bullish bias, below = bearish bias - Key edge: deviation from VWAP → often reverts = high-probability entry setup 7. Volatility Filter - when NOT to trade - Uses ATR (Average True Range) - Rule: only trade within the optimal volatility range

95,088 views

This account is 3 days old and already made $149,403 Nobody is watching this wallet and he made $30K only today His strategy is the most aggressive version of the Markov approach I've seen He doesn't enter at 50¢ or 70¢ He enters at 13¢–36¢ - the zone where the market treats the outcome as nearly impossible Then it resolves at 100¢ Fastest way to copy-trade him even with $10 using: Best closed trades: → BTC Up @ 21.4¢ - $28,814 · +368% → BTC Up @ 27.7¢ - $30,504 · +260% → BTC Down @ 33.5¢ - $34,587 · +195% He deploys $3K–$12K per position. At these returns, one win covers the next 5 entries The crowd sees 13¢ and reads: "won't happen." The Markov matrix reads the state transition and sees a committed direction Profile:

78,739 views

Claude Opus 4.7 + Polymarket - the best trading bot setup $388 → $93,000 in one month Bot scans Polymarket weather markets every 60 min Pulls 3 independent forecasts per city: ECMWF, HRRR, NOAA METAR Matches predictions to market buckets (14°C Seoul, 48–49°F Chicago) Buys YES at 5¢ when models agree, Kelly-sized Nightly: Opus reads journal & rewrites per-city rules Fastest way to copy-trade him even with $10 using: The Stack Hermes agent (body) + Claude Opus 4.7 (brain) + Hetzner $5 VPS + Telegram gateway + Visual Crossing API. Open-source repo: alteregoeth-ai/weatherbot. Funding – $10 min / $50 recommended in USDC.e – 2 POL for gas (~$1) – Approve 3 Polymarket contracts (CTF Exchange, Neg Risk, Router) Setup Time ~30 minutes. No code required. Just copy-paste 8 prompts into Hermes CLI Profile:

75,088 views

The first time I've seen a bot that turned $87 into $443,018 in just one month Nobody is watching this wallet and he makes $15.000+ every day 16,000 trades in 30 days = one prediction every 2.7 minutes, 24/7 His strategy: Trades 5-minute Up/Down windows across BTC, ETH, SOL, BNB and XRP simultaneously Enters at 73¢–96¢. Collects the spread to 100¢ momentum already committed - and loads just before resolution Fastest way to copy-trade him even with $10 using: Finds the window where momentum is locked That's a money printer with a Polymarket API key Profile:

44,633 views

This HFT bot is about to cross $1M profit trading 5-min BTC markets and making $10,000/day I traced his trades using Claude → backtested the strategy on 72M trades dataset the edge is that he runs both sides of the same window at once, then lets a probability model decide where the real money is that's the floor. the edge on top is directional: edge = fair_prob - market_price for each window the bot estimates the true probability BTC moves a direction in the next few minutes then fades the crowd when the price disagrees. that's why his fills run from 21¢ longshots to 99¢ near-locks across BTC and ETH Profile:

24,643 views

Chinese quant trader built Polymarket bot and turned $1.2K → $275,038 in one month using one secret He runs two separate engines simultaneously his edge is a Bayesian Prior Update before each session: P(UP | session) = P(UP) × P(futures_signal | UP) / P(futures_signal) engine 1-4-hour and session windows most quants ignore these. too slow, too much can go wrong he uses them because on a 4-hour window price has structure. futures market has already priced in direction. funding rate signals imbalance. open interest shows where the money is sitting when overnight BTC futures close with positive bias + funding rate negative → he bets UP on the 8AM-12PM window engine 2-5-minute windows same EV formula, different signal. no macro input. pure price state transitions when UP + DOWN < $1.00 → buy both sides, lock the arb gap when one side trades below 15¢ with trend confirmation → directional bet, size with Kelly the two engines don't compete. long windows = high conviction, large size. 5-minute windows = high frequency, smaller size, compounding Profile:

18,702 views

How you can build a Polymarket BTC trading engine with Claude to $1000/day print > RSI (context) + MACD (momentum) + Stochastic (timing) + EMA (structure) + OBV (confirmation) + VWAP (value) + Volatility (filter) = a bot that stops guessing and executes a framework Build v1: signal-based engine • Use RSI, ATR and cross-exchange price divergence to catch whale moves • Subscribe to Polymarket's order book WebSocket for real-time data • Minimize API calls. Let the WebSocket do the heavy lifting • This approach works around 95% of the time • The remaining 5% will cost you serious money if you hold to resolution Fastest way to copy-trade him even with $10 using: The edge No strategy wins forever Patterns only appear during specific windows of the day Your job: find the window. Trade only then AI alone won't write a strategy for you You need to train it on your own order book data first

How you can build a Polymarket BTC trading engine with Claude to $1000/day print > RSI (context) + MACD (momentum) + Stochastic (timing) + EMA (structure) + OBV (confirmation) + VWAP (value) + Volatility (filter) = a bot that stops guessing and executes a framework Build v1: signal-based engine • Use RSI, ATR and cross-exchange price divergence to catch whale moves • Subscribe to Polymarket's order book WebSocket for real-time data • Minimize API calls. Let the WebSocket do the heavy lifting • This approach works around 95% of the time • The remaining 5% will cost you serious money if you hold to resolution Fastest way to copy-trade him even with $10 using: The edge No strategy wins forever Patterns only appear during specific windows of the day Your job: find the window. Trade only then AI alone won't write a strategy for you You need to train it on your own order book data first

24,856 views

Ex-Citibank quant trader built system that turned him $2,217 → $199,881 in one month he's making $7,000–$10,000 every single day on autopilot 18,372 predictions, but no one knows he exists the system runs a complete-set delta model across BTC, ETH, SOL and XRP simultaneously: delta = P(UP) + P(DOWN) - 1.00 when delta 0 and trend confirms - size with Kelly, enter before the crowd adjusts right now he has 9 open positions across 4 assets in the same 5-minute window. running both engines at once. the arb leg covers downside, the directional leg prints on top most traders pick one direction and hope. he removed hope from the equation entirely Profile:

14,683 views

Whale bot printed $50k/week using Bayesian Update formula Start with prior from Markov state history, update on micro-signals in real time → posterior becomes your edge probability Formula: P(H|E) = [P(E|H) × P(H)] / P(E) (H = hypothesis like "BTC up in next 5 min", E = new evidence: order flow spike + vol cluster) Check him -> - Target ultra-short BTC/ETH/SOL/XRP Up/Down windows (5–15 min) - Layer evidence: net buys >2σ, on-chain tx acceleration, cross-asset momentum delta - Only enter if posterior shifts >10–15% from market price → high-conviction micro-edge - 92,722 predictions = ~300 trades/day compounding at scale Small belief updates × 300 trades/day × geometric compounding = $50k/week Merlin just shipped Whale Autotrade It tracks every profitable whale and detects the entry the moment it executes, mirrors it to your account instantly

Whale bot printed $50k/week using Bayesian Update formula Start with prior from Markov state history, update on micro-signals in real time → posterior becomes your edge probability Formula: P(H|E) = [P(E|H) × P(H)] / P(E) (H = hypothesis like "BTC up in next 5 min", E = new evidence: order flow spike + vol cluster) Check him -> - Target ultra-short BTC/ETH/SOL/XRP Up/Down windows (5–15 min) - Layer evidence: net buys >2σ, on-chain tx acceleration, cross-asset momentum delta - Only enter if posterior shifts >10–15% from market price → high-conviction micro-edge - 92,722 predictions = ~300 trades/day compounding at scale Small belief updates × 300 trades/day × geometric compounding = $50k/week Merlin just shipped Whale Autotrade It tracks every profitable whale and detects the entry the moment it executes, mirrors it to your account instantly

14,153 views

This Polymarket bot made $282,385 today Joined: Nov 2025, 17,866 predictions, Biggest win: $142,900 It trades tennis and football The edge: Polymarket lags behind by 30–90 seconds In that window: → sportsbook says 73% probability → Polymarket still shows 48.6¢ The bot enters at 48.6¢ The market catches up The bot exits at 93.5¢ Profit: 92.2% You should try to copy-trade him with a 0-second delay: The math per trade: Entry: 48.6¢ (implies 48.6% probability) Sportsbook real-time: 73% Gap: 24.4¢ per share EV = 0.73 × (1/0.486) − 1 = +0.502 $33,000 position → +$16,566 expected value Actual result: +$30,498 The model is working exactly as designed

This Polymarket bot made $282,385 today Joined: Nov 2025, 17,866 predictions, Biggest win: $142,900 It trades tennis and football The edge: Polymarket lags behind by 30–90 seconds In that window: → sportsbook says 73% probability → Polymarket still shows 48.6¢ The bot enters at 48.6¢ The market catches up The bot exits at 93.5¢ Profit: 92.2% You should try to copy-trade him with a 0-second delay: The math per trade: Entry: 48.6¢ (implies 48.6% probability) Sportsbook real-time: 73% Gap: 24.4¢ per share EV = 0.73 × (1/0.486) − 1 = +0.502 $33,000 position → +$16,566 expected value Actual result: +$30,498 The model is working exactly as designed

17,309 views

The Quant Who Turned Shannon's 77-Year-Old Formula Into +$555K in 30 Days He's not betting on crypto. He's compressing its uncertainty - one bit at a time His strategy: Shannon → Kelly → Crypto Up/Down The Formula D_KL(P ∥ Q) = Σ p(xᵢ) · log₂ [ p(xᵢ) / q(xᵢ) ] P = his model. Q = the market price. The result = how many bits of information he has that the market doesn't. That number IS the edge KL > 0.10 bits → Pro zone. Trade KL > 0.30 bits → Insane edge or bad math Every position is a binary, ultra-short crypto market: BTC, ETH, SOL, XRP - Up or Down on 5-min and hourly windows The same playground Shannon's math was built for Fastest way to copy-trade him even with $10 using: Kelly Sizing Position size scales with edge in bits: - High conviction (BTC Down 58.9¢): 2,999 shares - Medium (ETH Down 63.9¢): 900 shares - Long-tail (XRP Up 17.1¢): 158 shares Same recipe Thorp used to run 20 seasons with zero losing years While retail argues vibes on 5-min candles, this wallet runs KL-divergence and Kelly - 33,795 times in a row

The Quant Who Turned Shannon's 77-Year-Old Formula Into +$555K in 30 Days He's not betting on crypto. He's compressing its uncertainty - one bit at a time His strategy: Shannon → Kelly → Crypto Up/Down The Formula D_KL(P ∥ Q) = Σ p(xᵢ) · log₂ [ p(xᵢ) / q(xᵢ) ] P = his model. Q = the market price. The result = how many bits of information he has that the market doesn't. That number IS the edge KL > 0.10 bits → Pro zone. Trade KL > 0.30 bits → Insane edge or bad math Every position is a binary, ultra-short crypto market: BTC, ETH, SOL, XRP - Up or Down on 5-min and hourly windows The same playground Shannon's math was built for Fastest way to copy-trade him even with $10 using: Kelly Sizing Position size scales with edge in bits: - High conviction (BTC Down 58.9¢): 2,999 shares - Medium (ETH Down 63.9¢): 900 shares - Long-tail (XRP Up 17.1¢): 158 shares Same recipe Thorp used to run 20 seasons with zero losing years While retail argues vibes on 5-min candles, this wallet runs KL-divergence and Kelly - 33,795 times in a row

12,755 views

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

300 agents can find 1,000 sources. that still doesn’t make the research good the interesting part starts when those sources are forced to check each other. in K3 graph-native setup: 300 agents → 140 linked sources → 100% traceable claims → 9 contradictions surfaced one source confirms a number. another challenges it. a third connects the same entity from a different angle. so instead of: source → summary you get: source → entity → overlap → contradiction → verification → graph every new source becomes a node. if 3–4 independent sources support the same claim, confidence goes up. if 2 sources disagree, the contradiction gets flagged. if a claim has only 1 supporting source, it stays weak. a normal research workflow could have quietly shipped those 9 contradictions as facts. that’s why the breakthrough isn’t just 300 agents running in parallel. full breakdown in the article below.

300 agents can find 1,000 sources. that still doesn’t make the research good the interesting part starts when those sources are forced to check each other. in K3 graph-native setup: 300 agents → 140 linked sources → 100% traceable claims → 9 contradictions surfaced one source confirms a number. another challenges it. a third connects the same entity from a different angle. so instead of: source → summary you get: source → entity → overlap → contradiction → verification → graph every new source becomes a node. if 3–4 independent sources support the same claim, confidence goes up. if 2 sources disagree, the contradiction gets flagged. if a claim has only 1 supporting source, it stays weak. a normal research workflow could have quietly shipped those 9 contradictions as facts. that’s why the breakthrough isn’t just 300 agents running in parallel. full breakdown in the article below.

227,257 views • 17 days ago

Andrej Karpathy just said what no one wants to admit: "Your agent loop is generating garbage at scale and you can't see it." 1 output looks fine. 1,000 outputs = dead entropy, dead diversity. 90% of AI companies are building demos, not products. In a 20-minute conversation, Karpathy reveals the full reasoning stack

Andrej Karpathy just said what no one wants to admit: "Your agent loop is generating garbage at scale and you can't see it." 1 output looks fine. 1,000 outputs = dead entropy, dead diversity. 90% of AI companies are building demos, not products. In a 20-minute conversation, Karpathy reveals the full reasoning stack

312,765 views • 1 month ago

Anthropic Engineer Andrej Karpathy just broke the entire premise of modern AI: "Agents aren't magic. They're distillation at scale." RAG → Graph RAG → Memory → Multimodal Agent Graphs. 99.99% of your LLM's capacity is wasted on garbage data it never needed. In a 20-minute conversation, Karpathy reveals the full reasoning stack Worth more than any $500 AI course you've seen this year.

Anthropic Engineer Andrej Karpathy just broke the entire premise of modern AI: "Agents aren't magic. They're distillation at scale." RAG → Graph RAG → Memory → Multimodal Agent Graphs. 99.99% of your LLM's capacity is wasted on garbage data it never needed. In a 20-minute conversation, Karpathy reveals the full reasoning stack Worth more than any $500 AI course you've seen this year.

19,476 views • 3 days ago

Andrej Karpathy just reveales how LLMs actually thinks: "GPT-4 knows it failed. It just won't tell you unless you ask." >80% of GPT-4 errors are recoverable - the model already knows it screwed up. It has 80 transformer layers and spends the SAME compute on every single token as your brain In a 20-minute speach at Microsoft Build, Karpathy reveals the full psychology of LLMs. Worth more than any $500 prompting course you've seen on your timeline.

Andrej Karpathy just reveales how LLMs actually thinks: "GPT-4 knows it failed. It just won't tell you unless you ask." >80% of GPT-4 errors are recoverable - the model already knows it screwed up. It has 80 transformer layers and spends the SAME compute on every single token as your brain In a 20-minute speach at Microsoft Build, Karpathy reveals the full psychology of LLMs. Worth more than any $500 prompting course you've seen on your timeline.

163,114 views • 1 month ago

Andrej Karpathy in Stanford classroom said: "GPT is not a language model - It's a general-purpose computer" 90% of AI Twitter still thinks it's just autocomplete. It's not. >90% of the tools built on top of LLMs treat them like a smarter search engine. In this 40-minute lecture, Karpathy builds nanoGPT live 300 lines, reproduces GPT-2, explains every single decision. Worth more than any $500 deep learning bootcamp.

Andrej Karpathy in Stanford classroom said: "GPT is not a language model - It's a general-purpose computer" 90% of AI Twitter still thinks it's just autocomplete. It's not. >90% of the tools built on top of LLMs treat them like a smarter search engine. In this 40-minute lecture, Karpathy builds nanoGPT live 300 lines, reproduces GPT-2, explains every single decision. Worth more than any $500 deep learning bootcamp.

117,471 views • 1 month ago

300 agents without context engineering is just 300 ways to lose track of the answer. a serious agent system starts looking less like: 1 model → 1 prompt → 1 answer and more like: 1 planner → N parallel agents → verifier → merge → memory → human gate once you have 10, 50, or 300 agents, the number of possible relationships explodes. 100 entities = 4,950 possible pairwise relationships. 300 = 44,850. 1,000 = 499,500. that’s why “just give the model more context” stops working surprisingly fast. you need different agents seeing different slices of the graph. the architecture around the model is becoming the product. full article below ↓

300 agents without context engineering is just 300 ways to lose track of the answer. a serious agent system starts looking less like: 1 model → 1 prompt → 1 answer and more like: 1 planner → N parallel agents → verifier → merge → memory → human gate once you have 10, 50, or 300 agents, the number of possible relationships explodes. 100 entities = 4,950 possible pairwise relationships. 300 = 44,850. 1,000 = 499,500. that’s why “just give the model more context” stops working surprisingly fast. you need different agents seeing different slices of the graph. the architecture around the model is becoming the product. full article below ↓

32,708 views • 16 days ago

300 agents sounds powerful until you realize they can create 44,850 possible agent-to-agent relationships. that’s the real scaling problem. 1 agent searches 10 pages. 10 agents can search 100. 300 agents can search thousands in parallel. but more agents also means more: • duplicate findings • shared sources • hidden relationships and if every agent returns its own isolated report, you’re left with 300 piles to reconcile manually. that’s why the graph matters. instead of: 300 agents → 300 reports you get: 300 agents → shared entities → edges → clusters → one connected research map 100 nodes already create up to 4,950 pairwise relationships. more agents give you more information. the graph is what stops that information from becoming noise. full breakdown in the article below ↓

300 agents sounds powerful until you realize they can create 44,850 possible agent-to-agent relationships. that’s the real scaling problem. 1 agent searches 10 pages. 10 agents can search 100. 300 agents can search thousands in parallel. but more agents also means more: • duplicate findings • shared sources • hidden relationships and if every agent returns its own isolated report, you’re left with 300 piles to reconcile manually. that’s why the graph matters. instead of: 300 agents → 300 reports you get: 300 agents → shared entities → edges → clusters → one connected research map 100 nodes already create up to 4,950 pairwise relationships. more agents give you more information. the graph is what stops that information from becoming noise. full breakdown in the article below ↓

26,875 views • 14 days ago

300 agents just worked in parallel. the hard part starts when they’re done. The real unlock is the shared context graph. 100 entities already have 4,950 possible pairwise relationships. at 300, that jumps to 44,850. 300 agents → thousands of steps → shared entities → shared relationships → one connected memory layer Every run adds more structure instead of starting from zero. That means the next agent doesn’t just get more data. It gets a better map of how everything connects. That’s the part people are underestimating about multi-agent systems.

300 agents just worked in parallel. the hard part starts when they’re done. The real unlock is the shared context graph. 100 entities already have 4,950 possible pairwise relationships. at 300, that jumps to 44,850. 300 agents → thousands of steps → shared entities → shared relationships → one connected memory layer Every run adds more structure instead of starting from zero. That means the next agent doesn’t just get more data. It gets a better map of how everything connects. That’s the part people are underestimating about multi-agent systems.

34,737 views • 19 days ago

300 agents can still give you bad research. unless the sources can check each other. the interesting part is what happens when every source can see every other source. K3 swarm doesn't just dump research into 300 separate write-ups. • every source becomes a node. • shared entities become edges. • single-source claims stick out as weak. so instead of: `300 agents → 300 answers` you get: `300 agents → sources → links → one research graph` and that's where things start showing up that no individual agent was explicitly asked to find. the article's example ends with 140 linked sources, every claim traceable to its origin, and 9 contradictions surfaced instead of silently shipped as facts. esearch stops being a pile of tabs. it becomes a system where the sources check each other. Book this & read full breakdown below

300 agents can still give you bad research. unless the sources can check each other. the interesting part is what happens when every source can see every other source. K3 swarm doesn't just dump research into 300 separate write-ups. • every source becomes a node. • shared entities become edges. • single-source claims stick out as weak. so instead of: `300 agents → 300 answers` you get: `300 agents → sources → links → one research graph` and that's where things start showing up that no individual agent was explicitly asked to find. the article's example ends with 140 linked sources, every claim traceable to its origin, and 9 contradictions surfaced instead of silently shipped as facts. esearch stops being a pile of tabs. it becomes a system where the sources check each other. Book this & read full breakdown below

32,618 views • 18 days ago

300 agents without shared context is just parallelized confusion. All 300 agents are useless if the research dies inside 300 separate documents. the real unlock is what happens after the search. in K3 graph-native setup: 300 agents 140 linked sources 100% traceable claims 9 contradictions surfaced and every new source changes the graph. now imagine that happening across 8–10 active research clusters at once. instead of 300 isolated summaries, you get one shared structure: source → entity → relationship → evidence → contradiction → verification the math gets ugly fast. 100 entities = 4,950 possible pairwise relationships. that’s why “just run more agents” stops being enough. you need a graph that can keep track of what connects to what. full breakdown in the article below.

300 agents without shared context is just parallelized confusion. All 300 agents are useless if the research dies inside 300 separate documents. the real unlock is what happens after the search. in K3 graph-native setup: 300 agents 140 linked sources 100% traceable claims 9 contradictions surfaced and every new source changes the graph. now imagine that happening across 8–10 active research clusters at once. instead of 300 isolated summaries, you get one shared structure: source → entity → relationship → evidence → contradiction → verification the math gets ugly fast. 100 entities = 4,950 possible pairwise relationships. that’s why “just run more agents” stops being enough. you need a graph that can keep track of what connects to what. full breakdown in the article below.

15,089 views • 15 days ago

Andrej Karpathy dropped a lecture that makes every AI influencer feel stupid He just exposed the dirty secret of LLMs "90% of AI hype collapses when you realize even the builders can't explain what's inside." A 70B model is literally 2 files: parameters + 500 lines of C code >80% of AI safety work is patching holes in a sinking boat - Most people pay $1000 this knowledge. You’re getting it for free

Andrej Karpathy dropped a lecture that makes every AI influencer feel stupid He just exposed the dirty secret of LLMs "90% of AI hype collapses when you realize even the builders can't explain what's inside." A 70B model is literally 2 files: parameters + 500 lines of C code >80% of AI safety work is patching holes in a sinking boat - Most people pay $1000 this knowledge. You’re getting it for free

38,790 views • 1 month ago

Andrej Karpathy just broke the entire premise of modern AI: "The new computer isn't a chip - It's a language model running agent loops." Agents talking to agents -loops spawning sub-loops The stack being built RIGHT NOW: agent loop + memory + sub-agents + harness In a 20-minute talk, Karpathy explain the full reasoning stack

Andrej Karpathy just broke the entire premise of modern AI: "The new computer isn't a chip - It's a language model running agent loops." Agents talking to agents -loops spawning sub-loops The stack being built RIGHT NOW: agent loop + memory + sub-agents + harness In a 20-minute talk, Karpathy explain the full reasoning stack

31,924 views • 1 month ago

this 40 min MIT lecture on Markov Chains will teach you more about prediction markets than 5 years in trading it's 106 years old and wrote in 1906 > Google used it to build their search engine > quants use it to crush prediction markets > you can watch it for free tonight the algorithm that built the internet is the same one printing money on Polymarket Fastest way to copy-trade anyone even with $10 using:

this 40 min MIT lecture on Markov Chains will teach you more about prediction markets than 5 years in trading it's 106 years old and wrote in 1906 > Google used it to build their search engine > quants use it to crush prediction markets > you can watch it for free tonight the algorithm that built the internet is the same one printing money on Polymarket Fastest way to copy-trade anyone even with $10 using:

58,280 views • 5 months ago

Anthropic Engineer Andrej Karpathy: "LLM models writes code - and now it writes it better than you." Now LLMs write 70–90% of production code at top AI labs. Agents need the same thing: a loop that verifies its own output. No baseline = no benchmark = no improvement. In this 15-min talk, Karpathy breaks down how AI models actually works.

Anthropic Engineer Andrej Karpathy: "LLM models writes code - and now it writes it better than you." Now LLMs write 70–90% of production code at top AI labs. Agents need the same thing: a loop that verifies its own output. No baseline = no benchmark = no improvement. In this 15-min talk, Karpathy breaks down how AI models actually works.

11,402 views • 1 month ago

No more content to load