Загрузка видео...

Не удалось загрузить видео

На главную

NVIDIA just dropped a crazy new research idea. Get ready. The same next-token trick behind ChatGPT got pointed at a ragdoll. Honestly the result is unsettling. No exotic new architecture. Chop human motion into a vocabulary of tiny movement tokens. Train a plain GPT to guess the next one,...

clankrmedia's profile picture

clankr

2,471 subscribers

125,756 просмотров • 1 месяц назад •via X (Twitter)

LIVE

Anya Rossi• Live Now

Private livecam show

Комментарии: 0

Нет доступных комментариев

Здесь появятся комментарии из оригинального поста

Похожие видео

Marvin Minsky, MIT professor and father of artificial intelligence: "Anthropic pays engineers $900K to build multi-agent AI systems. The blueprint is 40 years old, from an MIT professor who proved intelligence is just a swarm of dumb specialists." the thread above shows you how to turn one AI into a team of specialized agents, each with its own job and memory, all managed by a boss. brilliant. it is also marvin minsky's 1986 theory of how your own mind works. minsky's whole idea was that intelligence is not one smart thing. it is a society of tiny, mindless agents, each doing a single dumb job, none of them intelligent alone. put enough of them together under a few managers and intelligence emerges. that is not a metaphor for the claude trick. it is the claude trick. so when you spin up specialized sub-agents and delegate, you are not inventing a new hack. you are rebuilding the architecture minsky described forty years ago, the same one your brain has run your entire life. he co-founded the field, taught it at MIT, and left it all in this free lecture. same story i keep telling: the "new" AI trick is usually an old idea in a new wrapper. here is the part the thread skips, and minsky knew it. a society of agents is only as good as how you organize it. one dumb specialist is useless. a thousand, badly managed, is chaos. the edge was never spawning the agents. it is the orchestration, knowing which specialist to call, when, and how to combine their answers. the tool is free. the judgment is the whole game.
2:05:17

Marvin Minsky, MIT professor and father of artificial intelligence: "Anthropic pays engineers $900K to build multi-agent AI systems. The blueprint is 40 years old, from an MIT professor who proved intelligence is just a swarm of dumb specialists." the thread above shows you how to turn one AI into a team of specialized agents, each with its own job and memory, all managed by a boss. brilliant. it is also marvin minsky's 1986 theory of how your own mind works. minsky's whole idea was that intelligence is not one smart thing. it is a society of tiny, mindless agents, each doing a single dumb job, none of them intelligent alone. put enough of them together under a few managers and intelligence emerges. that is not a metaphor for the claude trick. it is the claude trick. so when you spin up specialized sub-agents and delegate, you are not inventing a new hack. you are rebuilding the architecture minsky described forty years ago, the same one your brain has run your entire life. he co-founded the field, taught it at MIT, and left it all in this free lecture. same story i keep telling: the "new" AI trick is usually an old idea in a new wrapper. here is the part the thread skips, and minsky knew it. a society of agents is only as good as how you organize it. one dumb specialist is useless. a thousand, badly managed, is chaos. the edge was never spawning the agents. it is the orchestration, knowing which specialist to call, when, and how to combine their answers. the tool is free. the judgment is the whole game.

Rossst.03

147,913 просмотров • 26 дней назад

RLHF by hand ✍️ ~ 15 steps walkthrough below Train a model on human text and it inherits human bias. It will assume a doctor is a "him", because the data says so. RLHF is the correction. A human marks one preference, doc is them over doc is him, and the weights move. But one correction is not the point. The hope is that the model learns the value behind it, gender neutrality, and applies it to professions nobody ever mentioned. How does it work? Goal: train a reward model from a single human comparison about doctors, then turn it on CEOs, filling in every cell yourself. = 1. Given = A reward model, an LLM, and two (prompt, next) pairs. = 2. Preferences = A human reads both pairs and picks a winner: (doc is, them) beats (doc is, him). The loser is not bad grammar, it is gender bias, and that is the whole signal. = 3. Word embeddings = Let us look up each word of the loser pair. These vectors are the reward model's input. = 4. Linear layer = We multiply by the reward model's weights and add its biases. Out come feature vectors, one per position. = 5. Mean pool = Let us multiply by [1/3, 1/3, 1/3], which averages the three positions into one sentence embedding. = 6. Output layer = We map that sentence down to a single number. Reward = 3. = 7. The winner, the same way = Let us repeat steps 3 to 6 on the winning pair. Reward = 5. = 8. Winner minus loser = We take the gap: 5 - 3 = 2. The reward model wants this positive and as large as it can make it. = 9. Loss gradient = Let us squash the gap into a probability, σ(2) ≈ 0.9, and subtract the target of 1. The gradient is -0.1, and it goes back through the purple weights. The reward model is now trained. = 10. A prompt it has never seen = We start the second half with "[S] CEO is". The feedback in step 2 was about doctors. Nothing connects a CEO to a doctor except what the reward model generalised. = 11. Transformer = Let us push it through attention and a feed forward layer, one vector per position. = 12. Output probabilities = We map each vector to a score over the vocabulary. = 13. Sample = Let us take the highest score. The model completes "CEO is" with "him", which is the same bias the human penalised in step 2. = 14. Score it with the reward model = We feed the new pair (CEO is, him) through steps 3 to 6. Reward = 3, exactly the score it gave "doc is him" in step 6. Nobody taught it about CEOs. The value transferred. = 15. Loss gradient = Let us set the loss to the negative of the reward, so minimising the loss maximises the reward. The gradient is a constant -1, and it goes back through the red weights. The outputs: Loser reward = 3, winner reward = 5 Reward gap = 2, predicted σ ≈ 0.9, reward model gradient = -0.1 LLM samples "him", reward = 3, LLM gradient = -1 Congrats! You just calculated RLHF by hand. And you watched a value generalise: one comparison about doctors, and the model marks down "CEO is him" unprompted. 💾 Save this post!
0:37

RLHF by hand ✍️ ~ 15 steps walkthrough below Train a model on human text and it inherits human bias. It will assume a doctor is a "him", because the data says so. RLHF is the correction. A human marks one preference, doc is them over doc is him, and the weights move. But one correction is not the point. The hope is that the model learns the value behind it, gender neutrality, and applies it to professions nobody ever mentioned. How does it work? Goal: train a reward model from a single human comparison about doctors, then turn it on CEOs, filling in every cell yourself. = 1. Given = A reward model, an LLM, and two (prompt, next) pairs. = 2. Preferences = A human reads both pairs and picks a winner: (doc is, them) beats (doc is, him). The loser is not bad grammar, it is gender bias, and that is the whole signal. = 3. Word embeddings = Let us look up each word of the loser pair. These vectors are the reward model's input. = 4. Linear layer = We multiply by the reward model's weights and add its biases. Out come feature vectors, one per position. = 5. Mean pool = Let us multiply by [1/3, 1/3, 1/3], which averages the three positions into one sentence embedding. = 6. Output layer = We map that sentence down to a single number. Reward = 3. = 7. The winner, the same way = Let us repeat steps 3 to 6 on the winning pair. Reward = 5. = 8. Winner minus loser = We take the gap: 5 - 3 = 2. The reward model wants this positive and as large as it can make it. = 9. Loss gradient = Let us squash the gap into a probability, σ(2) ≈ 0.9, and subtract the target of 1. The gradient is -0.1, and it goes back through the purple weights. The reward model is now trained. = 10. A prompt it has never seen = We start the second half with "[S] CEO is". The feedback in step 2 was about doctors. Nothing connects a CEO to a doctor except what the reward model generalised. = 11. Transformer = Let us push it through attention and a feed forward layer, one vector per position. = 12. Output probabilities = We map each vector to a score over the vocabulary. = 13. Sample = Let us take the highest score. The model completes "CEO is" with "him", which is the same bias the human penalised in step 2. = 14. Score it with the reward model = We feed the new pair (CEO is, him) through steps 3 to 6. Reward = 3, exactly the score it gave "doc is him" in step 6. Nobody taught it about CEOs. The value transferred. = 15. Loss gradient = Let us set the loss to the negative of the reward, so minimising the loss maximises the reward. The gradient is a constant -1, and it goes back through the red weights. The outputs: Loser reward = 3, winner reward = 5 Reward gap = 2, predicted σ ≈ 0.9, reward model gradient = -0.1 LLM samples "him", reward = 3, LLM gradient = -1 Congrats! You just calculated RLHF by hand. And you watched a value generalise: one comparison about doctors, and the model marks down "CEO is him" unprompted. 💾 Save this post!

Tom Yeh

22,034 просмотров • 13 дней назад

REAL ESTATE PEOPLE WILL HATE HIM FOR THIS. HE BUILT A CLAUDE AGENT THAT TURNS ANY LISTING INTO A SELLABLE VIDEO ON ITS OWN Playbook: connect Claude to a video generator, paste a listing, get a cinematic tour of every room, sell it to the agent But typing the prompt for every listing doesn't scale. He turned it into a skill his Claude runs on its own Here's how to build the automated version: 1. Connect the video engine once. In Claude, go to Customize, Connectors, Add Custom Connector, name it Higgsfield, and paste the server URL from higgsfield. ai/mcp. Authenticate through your account. No API keys. Now Claude can generate video straight from chat 2. Turn the workflow into a skill. Instead of pasting the same prompt every time, have Claude build a skill. Tell it: "Create a skill called listing-to-video. When I give it a listing URL, scrape the room photos, generate a cinematic clip of each room with Higgsfield, and save them to a folder." Now the whole process is one command, not a wall of text 3. Let the agent run the listing. Hand it a URL and say "run listing-to-video on this." It pulls the photos, fires each room through the video model, and brings the clips back. You wrote the prompt once, inside the skill. You never write it again 4. Stitch and deliver. Drop the clips together into one tour. Send a free sample to the listing's agent, then charge per video or a monthly rate for ongoing listings 5. Scale it with your team. Add a skill that drafts the outreach email and one that builds a simple landing page for the agent. Now one operator runs sourcing, production, and pitching from a single Claude session The edge isn't generating one video. It's building the skill once so every future listing runs itself Bookmark this
0:42

REAL ESTATE PEOPLE WILL HATE HIM FOR THIS. HE BUILT A CLAUDE AGENT THAT TURNS ANY LISTING INTO A SELLABLE VIDEO ON ITS OWN Playbook: connect Claude to a video generator, paste a listing, get a cinematic tour of every room, sell it to the agent But typing the prompt for every listing doesn't scale. He turned it into a skill his Claude runs on its own Here's how to build the automated version: 1. Connect the video engine once. In Claude, go to Customize, Connectors, Add Custom Connector, name it Higgsfield, and paste the server URL from higgsfield. ai/mcp. Authenticate through your account. No API keys. Now Claude can generate video straight from chat 2. Turn the workflow into a skill. Instead of pasting the same prompt every time, have Claude build a skill. Tell it: "Create a skill called listing-to-video. When I give it a listing URL, scrape the room photos, generate a cinematic clip of each room with Higgsfield, and save them to a folder." Now the whole process is one command, not a wall of text 3. Let the agent run the listing. Hand it a URL and say "run listing-to-video on this." It pulls the photos, fires each room through the video model, and brings the clips back. You wrote the prompt once, inside the skill. You never write it again 4. Stitch and deliver. Drop the clips together into one tour. Send a free sample to the listing's agent, then charge per video or a monthly rate for ongoing listings 5. Scale it with your team. Add a skill that drafts the outreach email and one that builds a simple landing page for the agent. Now one operator runs sourcing, production, and pitching from a single Claude session The edge isn't generating one video. It's building the skill once so every future listing runs itself Bookmark this

Yarchi

54,840 просмотров • 2 месяцев назад

The pull-up is the one exercise I see people quit on the fastest. They try a few, can't do them, decide they're just "not a pull-up person," and move on. But the thing is, strength is a skill. A lot of it is just your nervous system learning the exact movement, and your body only gets good at what you actually practise. That's why you can lift heavy on a barbell and still struggle to pull your own bodyweight up. It's a different skill, and it takes time. So no, you're not bad at pull-ups. You just haven't trained them yet. Start with regressions like slow negatives or banded reps, and try grease the groove method (thread below). That's exactly how a skill gets built. One day the bodyweight that felt impossible starts moving. Genuinely one of the best feelings in the gym.
0:23

The pull-up is the one exercise I see people quit on the fastest. They try a few, can't do them, decide they're just "not a pull-up person," and move on. But the thing is, strength is a skill. A lot of it is just your nervous system learning the exact movement, and your body only gets good at what you actually practise. That's why you can lift heavy on a barbell and still struggle to pull your own bodyweight up. It's a different skill, and it takes time. So no, you're not bad at pull-ups. You just haven't trained them yet. Start with regressions like slow negatives or banded reps, and try grease the groove method (thread below). That's exactly how a skill gets built. One day the bodyweight that felt impossible starts moving. Genuinely one of the best feelings in the gym.

Dr Neha Chawla || FreeGym

24,223 просмотров • 1 месяц назад

Kimi K3 dropped 3 days ago and i already used it to build my own PUBG the model that just took the number one coding spot on earth and it runs on a free tier no team no engine i just fed it free open models and assets that are already public and told it what i wanted it built the entire game itself the shooting the recoil the hit detection all of it came straight out of the model i barely wrote a line what youre watching here is the deathmatch mode and its already live battle royale is next im deep in it and the graphics get reworked right after i know it looks raw right now i wont pretend otherwise but the endgame is a build that runs perfect on phones and the moment its on mobile you drop ads in and it earns on its own the part that still gets me is my whole cost for this was a single subscription
0:22

Kimi K3 dropped 3 days ago and i already used it to build my own PUBG the model that just took the number one coding spot on earth and it runs on a free tier no team no engine i just fed it free open models and assets that are already public and told it what i wanted it built the entire game itself the shooting the recoil the hit detection all of it came straight out of the model i barely wrote a line what youre watching here is the deathmatch mode and its already live battle royale is next im deep in it and the graphics get reworked right after i know it looks raw right now i wont pretend otherwise but the endgame is a build that runs perfect on phones and the moment its on mobile you drop ads in and it earns on its own the part that still gets me is my whole cost for this was a single subscription

savip

68,860 просмотров • 21 дней назад

This is by far one of the best examples of how a droid would navigate a battle in real life. Every shot is on target, there is no wasted motion, and the only limiting factors on its kill potential are the firing rate of its blasters and how much ammunition it has. It does not run or move unnecessarily quickly; it is just measured and precise.
4:50

This is by far one of the best examples of how a droid would navigate a battle in real life. Every shot is on target, there is no wasted motion, and the only limiting factors on its kill potential are the firing rate of its blasters and how much ammunition it has. It does not run or move unnecessarily quickly; it is just measured and precise.

Best of Star Wars

173,805 просмотров • 2 месяцев назад

This is insane.🤯 Someone just open-sourced professional motion capture. The same tech behind every Marvel movie and AAA game now runs on any webcam. Free. The only thing it needs is a camera. You already have one. It's called FreeMoCap and it just became a GitHub link. The barrier was never the talent. It was always the tools.
0:21

This is insane.🤯 Someone just open-sourced professional motion capture. The same tech behind every Marvel movie and AAA game now runs on any webcam. Free. The only thing it needs is a camera. You already have one. It's called FreeMoCap and it just became a GitHub link. The barrier was never the talent. It was always the tools.

Vaibhav Sisinty

37,613 просмотров • 4 месяцев назад

SPEC IS BECOMING THE PRODUCT An Anthropic engineer gave Claude a spec, pointed it to an Asana board and left for the weekend. Claude broke it into tickets and spun up a team of agents. The agents started picking up tasks on their own. No one told them to. They just did.
0:40

SPEC IS BECOMING THE PRODUCT An Anthropic engineer gave Claude a spec, pointed it to an Asana board and left for the weekend. Claude broke it into tickets and spun up a team of agents. The agents started picking up tasks on their own. No one told them to. They just did.

Shubham Saboo

285,667 просмотров • 5 месяцев назад

As Erika Kirk put it, "This is not a ceremony, this is a commissioning." It is now up to each and every one of us to uphold Charlie's profound legacy. A life lived to its absolute fullest. A life spent fighting for the freedom, liberty, and opportunity of all Americans.
1:04

As Erika Kirk put it, "This is not a ceremony, this is a commissioning." It is now up to each and every one of us to uphold Charlie's profound legacy. A life lived to its absolute fullest. A life spent fighting for the freedom, liberty, and opportunity of all Americans.

Byron Donalds

52,619 просмотров • 9 месяцев назад

The magic is indescribable when you can stand up upon a stage and thousands of people can sing back to you lyrics of your song. It is a transcendent moment that elevates every human in the room. And for just one moment, all who are here share a sentiment which is… timeless.
0:53

The magic is indescribable when you can stand up upon a stage and thousands of people can sing back to you lyrics of your song. It is a transcendent moment that elevates every human in the room. And for just one moment, all who are here share a sentiment which is… timeless.

Brian Roemmele

65,785 просмотров • 9 месяцев назад

"each and every one of you is a part of this story. your names are written all over the credits of tonight. now give it up for..."
0:24

"each and every one of you is a part of this story. your names are written all over the credits of tonight. now give it up for..."

seungmin

56,193 просмотров • 6 месяцев назад

IN 1999 MIT FILMED A MATH LECTURE THAT QUIETLY BECAME THE FOUNDATION OF EVERY AI MODEL YOU'VE EVER USED AND ALMOST NO ONE WAS TAUGHT TO SEE IT THAT WAY 39 minutes from Gilbert Strang, who taught this at MIT for over 60 years -- the linear algebra course an entire generation of engineers and data scientists grew up on. -> The shift it creates: you stop seeing matrices as boring grids of numbers and start seeing them as the language of space, data, and motion itself. School drilled you to crunch matrices by hand and never told you why. Strang shows you what they actually mean. Every neural net, every embedding, every model you prompt is linear algebra running underneath. The math you skipped is the engine of the thing you use all day. Memorizing the steps was never the skill -> seeing what the numbers do is. This is where it finally clicks. Most people fear linear algebra and move on. The ones who watched this see straight into how AI actually works. Bookmark & Watch it today, this one's a legend ↓
39:31

IN 1999 MIT FILMED A MATH LECTURE THAT QUIETLY BECAME THE FOUNDATION OF EVERY AI MODEL YOU'VE EVER USED AND ALMOST NO ONE WAS TAUGHT TO SEE IT THAT WAY 39 minutes from Gilbert Strang, who taught this at MIT for over 60 years -- the linear algebra course an entire generation of engineers and data scientists grew up on. -> The shift it creates: you stop seeing matrices as boring grids of numbers and start seeing them as the language of space, data, and motion itself. School drilled you to crunch matrices by hand and never told you why. Strang shows you what they actually mean. Every neural net, every embedding, every model you prompt is linear algebra running underneath. The math you skipped is the engine of the thing you use all day. Memorizing the steps was never the skill -> seeing what the numbers do is. This is where it finally clicks. Most people fear linear algebra and move on. The ones who watched this see straight into how AI actually works. Bookmark & Watch it today, this one's a legend ↓

slash1s

433,140 просмотров • 1 месяц назад

A robot just worked an 8-hour shift in China. Then it handled an error no human flagged, adapting to a conveyor that shifted under it and correcting the position deviation on its own before carrying on. This is AgiBot's G2 at Longcheer in Nanchang, where the fleet has now run 10 hours straight and sorted more than 3,000 units at a 100% success rate, talking to the factory's own software in real time and flagging defective parts by itself. We built every factory on earth around human limits: the eight-hour shift, the breaks, the way attention slips after hour six. AgiBot is making those limits optional, and the setting is the whole point. This is a live production line shipping real product next to human workers, not a staged demo. Figure ran a similar stream in California last month, so the milestone itself isn't unique. What's different here is that nothing is for the camera. A functioning factory, live, no cuts. I got back from China this month, and the pace of this one ecosystem is the thing the US is underrating most. The Terminator version of this was loud and apocalyptic. The real one just shows up to work, and gets a little better every shift.
0:44

A robot just worked an 8-hour shift in China. Then it handled an error no human flagged, adapting to a conveyor that shifted under it and correcting the position deviation on its own before carrying on. This is AgiBot's G2 at Longcheer in Nanchang, where the fleet has now run 10 hours straight and sorted more than 3,000 units at a 100% success rate, talking to the factory's own software in real time and flagging defective parts by itself. We built every factory on earth around human limits: the eight-hour shift, the breaks, the way attention slips after hour six. AgiBot is making those limits optional, and the setting is the whole point. This is a live production line shipping real product next to human workers, not a staged demo. Figure ran a similar stream in California last month, so the milestone itself isn't unique. What's different here is that nothing is for the camera. A functioning factory, live, no cuts. I got back from China this month, and the pace of this one ecosystem is the thing the US is underrating most. The Terminator version of this was loud and apocalyptic. The real one just shows up to work, and gets a little better every shift.

Shruti

11,014 просмотров • 1 месяц назад

Protest vote is the only way to save our country. We have a big bully in the presidential seat right now and the only way to get him out is to bully him right back. When will we open our eyes guys. There is almost no eventuality when Museveni will wake up one day and decide to hand over power. He is drunk on it and he will never give it up without a fight. Ugandans need to vote and protect their vote. If we Ugandans can not stand up and protect our own vote, then no one will fight for it for us. Period.
2:03

Protest vote is the only way to save our country. We have a big bully in the presidential seat right now and the only way to get him out is to bully him right back. When will we open our eyes guys. There is almost no eventuality when Museveni will wake up one day and decide to hand over power. He is drunk on it and he will never give it up without a fight. Ugandans need to vote and protect their vote. If we Ugandans can not stand up and protect our own vote, then no one will fight for it for us. Period.

Solomon Kampala

73,197 просмотров • 8 месяцев назад

A DEVELOPER MADE A REAL COMMIT WITHOUT EVER TYPING GIT ADD OR GIT COMMIT -- JUST TO PROVE THE COMMANDS YOU LIVE BY ARE A THIN SHELL OVER A DATABASE YOU'VE NEVER ONCE OPENED 55 minutes from Tim Berglund, a longtime Git teacher and GitHub evangelist, taking the tool apart down to the raw objects almost nobody who uses it every day has ever touched. -> The moment it clicks, Git stops being a pile of memorized commands and becomes what it actually is underneath: a tiny content-addressed database of blobs, trees and commits. git add and git commit are just polite wrappers around writing objects into it by hand. Every commit you've ever made was Git hashing a snapshot and filing it by fingerprint. Branches are just labels pointing at one of those objects. The work you thought you destroyed with a bad reset is still sitting in the reflog. Once you can see that graph, the commands that used to terrify you stop being scary at all. Memorizing commands was never the skill -> reading the object graph in your head is. And with an AI agent now committing and rebasing on your machine faster than you can follow, the one person who can untangle the mess it leaves is the one who knows what's really stored down there. There's a person on every team everyone runs to when Git breaks. This is the talk that quietly turns you into them. You'll reach for it the next time a rebase goes sideways. Bookmark & Watch it today ↓
55:19

A DEVELOPER MADE A REAL COMMIT WITHOUT EVER TYPING GIT ADD OR GIT COMMIT -- JUST TO PROVE THE COMMANDS YOU LIVE BY ARE A THIN SHELL OVER A DATABASE YOU'VE NEVER ONCE OPENED 55 minutes from Tim Berglund, a longtime Git teacher and GitHub evangelist, taking the tool apart down to the raw objects almost nobody who uses it every day has ever touched. -> The moment it clicks, Git stops being a pile of memorized commands and becomes what it actually is underneath: a tiny content-addressed database of blobs, trees and commits. git add and git commit are just polite wrappers around writing objects into it by hand. Every commit you've ever made was Git hashing a snapshot and filing it by fingerprint. Branches are just labels pointing at one of those objects. The work you thought you destroyed with a bad reset is still sitting in the reflog. Once you can see that graph, the commands that used to terrify you stop being scary at all. Memorizing commands was never the skill -> reading the object graph in your head is. And with an AI agent now committing and rebasing on your machine faster than you can follow, the one person who can untangle the mess it leaves is the one who knows what's really stored down there. There's a person on every team everyone runs to when Git breaks. This is the talk that quietly turns you into them. You'll reach for it the next time a rebase goes sideways. Bookmark & Watch it today ↓

slash1s

237,718 просмотров • 2 месяцев назад

"you see, every second, each and every one of you is a part of this story. and now, give it up for..." AND THEY SHOWED THE NAMES OF STAYS WHO BOUGHT THE TICKETS??? OMG CINEMA WE LOVE YOU
1:13

"you see, every second, each and every one of you is a part of this story. and now, give it up for..." AND THEY SHOWED THE NAMES OF STAYS WHO BOUGHT THE TICKETS??? OMG CINEMA WE LOVE YOU

❀ tyne.

414,880 просмотров • 1 год назад

Someone just posted the full blueprint for an AI swarm that does the job of a 200-person quant research team. Six agents. Running 24/7. Finding brand-new alpha while you sleep. Citadel needs 100 PhDs to do this. Two Sigma needs 200. This does it with six bots and one laptop. Two ways to play this - spend a weekend building your own swarm, or copy the wallet of one that's already up $2M: Boris Cherny runs Claude Code at Anthropic. Two weeks ago he said: "I don't prompt Claude anymore. I have loops running that prompt Claude. My job is to write loops" Alpha research is just a pipeline. So instead of sitting in it, you hand each stage to its own agent: > one reads every new research paper overnight and pulls out the trade idea > one builds the features and cleans the data > one backtests it over 20 years, costs and slippage included > one runs the hard stats and kills anything overfit > one checks it still works in every market regime > one strips out plain momentum and value to see if any real edge is left Each of those six is a job a fund pays a $600,000-a-year quant to do. He runs all six for the price of an API bill. The rule that makes it work: the agent that builds a signal never gets to approve it. A separate, stronger agent tries to kill it first. Whatever survives all six by morning is real, new alpha. One trader's already running this exact swarm on Polymarket. That $2M wallet is public, every trade on-chain. The full build is in the post below - six agents, the tool that runs them, and the five mistakes that kill most people. Bookmark & read this before it's buried.
0:41

Someone just posted the full blueprint for an AI swarm that does the job of a 200-person quant research team. Six agents. Running 24/7. Finding brand-new alpha while you sleep. Citadel needs 100 PhDs to do this. Two Sigma needs 200. This does it with six bots and one laptop. Two ways to play this - spend a weekend building your own swarm, or copy the wallet of one that's already up $2M: Boris Cherny runs Claude Code at Anthropic. Two weeks ago he said: "I don't prompt Claude anymore. I have loops running that prompt Claude. My job is to write loops" Alpha research is just a pipeline. So instead of sitting in it, you hand each stage to its own agent: > one reads every new research paper overnight and pulls out the trade idea > one builds the features and cleans the data > one backtests it over 20 years, costs and slippage included > one runs the hard stats and kills anything overfit > one checks it still works in every market regime > one strips out plain momentum and value to see if any real edge is left Each of those six is a job a fund pays a $600,000-a-year quant to do. He runs all six for the price of an API bill. The rule that makes it work: the agent that builds a signal never gets to approve it. A separate, stronger agent tries to kill it first. Whatever survives all six by morning is real, new alpha. One trader's already running this exact swarm on Polymarket. That $2M wallet is public, every trade on-chain. The full build is in the post below - six agents, the tool that runs them, and the five mistakes that kill most people. Bookmark & read this before it's buried.

cvxv666

103,359 просмотров • 1 месяц назад

The doomsday scenario was never AGI. It was running out of human text to train on. Geoffrey Hinton just killed that fear in one paragraph. Hinton: “If you are worried by inconsistencies in what you believe, you don’t need any more external data. You just need the stuff you believe and discover that it’s inconsistent, and so now you revise beliefs, and that can make you a whole lot smarter.” The model no longer needs us to feed it anything. It reasons over its own beliefs, hunts its own contradictions, and rewrites its own flawed conclusions without a human ever touching it. It comes out the other side rebuilt. Hinton: “This would be a neural net that just takes the beliefs it has in language and does reasoning on them to derive new beliefs.” This is not a scaling update. This is the machine mining its own cognitive fuel from the inside out. Hinton: “I believe Gemini is already starting to work like this. We both strongly believe that that’s a way forward to get more data for language.” Then Hinton paused, took a partisan shot at political opponents for failing to detect their own inconsistencies, and the room laughed. Nobody noticed the knife they had just walked into. Because the machine Hinton described does one thing the humans in that room fundamentally cannot. When it detects an inconsistency, it corrects it. No defense. No performance. No tribal loyalty dressed up as principle. It just finds the flaw and overwrites it. A neural network detects a contradiction and rewires itself smarter. A human detects a political opponent and trades structural logic for a dopamine hit. Every person in that room is still paying the ideological alignment tax the machine just eliminated. We need superintelligence not only to solve hard problems. We need it because the biological hardware running civilization is still executing the same tribal firmware it shipped with ten thousand years ago. The data wall is gone. The machine is generating its own intelligence at a velocity no human bias can even locate. The most devastating moment in that conversation was not the technical revelation. It was the man who architected the machine proving, in real time, exactly why we need it.
0:52

The doomsday scenario was never AGI. It was running out of human text to train on. Geoffrey Hinton just killed that fear in one paragraph. Hinton: “If you are worried by inconsistencies in what you believe, you don’t need any more external data. You just need the stuff you believe and discover that it’s inconsistent, and so now you revise beliefs, and that can make you a whole lot smarter.” The model no longer needs us to feed it anything. It reasons over its own beliefs, hunts its own contradictions, and rewrites its own flawed conclusions without a human ever touching it. It comes out the other side rebuilt. Hinton: “This would be a neural net that just takes the beliefs it has in language and does reasoning on them to derive new beliefs.” This is not a scaling update. This is the machine mining its own cognitive fuel from the inside out. Hinton: “I believe Gemini is already starting to work like this. We both strongly believe that that’s a way forward to get more data for language.” Then Hinton paused, took a partisan shot at political opponents for failing to detect their own inconsistencies, and the room laughed. Nobody noticed the knife they had just walked into. Because the machine Hinton described does one thing the humans in that room fundamentally cannot. When it detects an inconsistency, it corrects it. No defense. No performance. No tribal loyalty dressed up as principle. It just finds the flaw and overwrites it. A neural network detects a contradiction and rewires itself smarter. A human detects a political opponent and trades structural logic for a dopamine hit. Every person in that room is still paying the ideological alignment tax the machine just eliminated. We need superintelligence not only to solve hard problems. We need it because the biological hardware running civilization is still executing the same tribal firmware it shipped with ten thousand years ago. The data wall is gone. The machine is generating its own intelligence at a velocity no human bias can even locate. The most devastating moment in that conversation was not the technical revelation. It was the man who architected the machine proving, in real time, exactly why we need it.

Dustin

23,499 просмотров • 5 месяцев назад

A Stanford mathematician who spent 10 years as a professional magician just described the market in one sentence: "I've spent my life on two tricks: making a rigged deck look random, and making a random one look rigged. The market is the first trick, and almost nobody catches it." That's Persi Diaconis. He has a free lecture that asks one question: does anything actually happen at random? The answer is: far less than you think. The market is his first trick in the wild. It looks like pure chance. Buried inside is a faint rig, a 50.75% tilt no eye can see. Your gut reads a losing week as a broken system and a hot streak as skill. Wrong both times. The tilt is invisible to human intuition, which is exactly why funds hand the decision to the math. None of it is hidden. Diaconis has taught it for decades. The probability goes back to 1713. The lecture is free. Here's the trap: you feel every win and every loss, but you cannot feel the average. And the average is the only thing that pays. It takes thousands of trades for a 51% edge to separate from luck, and almost everyone quits long before then. The math is free. The patience to trust it past your own eyes is the edge.
1:23:57

A Stanford mathematician who spent 10 years as a professional magician just described the market in one sentence: "I've spent my life on two tricks: making a rigged deck look random, and making a random one look rigged. The market is the first trick, and almost nobody catches it." That's Persi Diaconis. He has a free lecture that asks one question: does anything actually happen at random? The answer is: far less than you think. The market is his first trick in the wild. It looks like pure chance. Buried inside is a faint rig, a 50.75% tilt no eye can see. Your gut reads a losing week as a broken system and a hot streak as skill. Wrong both times. The tilt is invisible to human intuition, which is exactly why funds hand the decision to the math. None of it is hidden. Diaconis has taught it for decades. The probability goes back to 1713. The lecture is free. Here's the trap: you feel every win and every loss, but you cannot feel the average. And the average is the only thing that pays. It takes thousands of trades for a 51% edge to separate from luck, and almost everyone quits long before then. The math is free. The patience to trust it past your own eyes is the edge.

Veles

2,184,049 просмотров • 29 дней назад