Loading video...

Video Failed to Load

There was a problem loading this video. This could be due to a temporary network issue or the video might be unavailable.

You don't need a trillion-parameter model that understands physics to answer a customer support call. But that's not even the main reason we build small models. The real one: LLMs store everything. Literally everything all of it sitting in the weights. Humans don't work that way, and your agent... show more

Sudarshan Kamath

22,292 subscribers

19,999 views • 3 months ago •via X (Twitter)

Science & Technology

Anya Rossi• Live Now

Private livecam show

5 Comments

shriraj3 months ago

Yo @0xSarthak lookalike

Arun Sharma3 months ago

Strongly agree. But you can't compete with frontier LLMs with small models + postgres/sqlite. The storage needs to be rethought and redesigned.

Pratik Poddar3 months ago

Very cool. Exciting stuff!

John Wei3 months ago

yeah smarter systems will probably beat bigger models

Smit B Patel3 months ago

The parameter variables store all the essence of the data!

Related Videos

Geoffrey Hinton says LLMs are immortal. Destroy the hardware, save the weights. Rebuild them later, and “the very same being comes back to life.” But humans don't work that way. We're bound to our substrate. Not even Kurzweil gets a backup.

Geoffrey Hinton says LLMs are immortal. Destroy the hardware, save the weights. Rebuild them later, and “the very same being comes back to life.” But humans don't work that way. We're bound to our substrate. Not even Kurzweil gets a backup.

vitrupo

323,413 views • 1 year ago

Satya Nadella on Microsoft's CapEx and CoPilot: "We are calibrating our CapEx in such a way that we don't want to build for 1 or 2 customers, we want to build for the long tail because I think that's the most important. If you are a hyperscaler, you're not a supplier to 2 model companies, that's not a business. You have to sort of build a system that is great for lots of third parties and our own 1P. In that context, we are pretty thrilled with the progress we are making even with CoPilot if you look at the subscriber number we gave. These are real enterprises using it for real workflows." Via The All-In Podcast $MSFT $GOOGL $AMZN $META

Satya Nadella on Microsoft's CapEx and CoPilot: "We are calibrating our CapEx in such a way that we don't want to build for 1 or 2 customers, we want to build for the long tail because I think that's the most important. If you are a hyperscaler, you're not a supplier to 2 model companies, that's not a business. You have to sort of build a system that is great for lots of third parties and our own 1P. In that context, we are pretty thrilled with the progress we are making even with CoPilot if you look at the subscriber number we gave. These are real enterprises using it for real workflows." Via The All-In Podcast $MSFT $GOOGL $AMZN $META

P Equity Research 📰

33,817 views • 1 month ago

WebMCP makes sites agent-native. The industries that need it first: - any complicated web store - SaaS admin boards and consoles - analytics and marketing tools - read-only helpers in insurance or banking self-service That last one is the unlock for regulated industries. A few tools that read browser state, touch nothing, and either point the user to the right screen or pull the right info for their agent. But as a founder, you don't need any of that for your customers right away. Build it in-house first. Your internal tools, your own team, your own workflow. That's where the potential is.

WebMCP makes sites agent-native. The industries that need it first: - any complicated web store - SaaS admin boards and consoles - analytics and marketing tools - read-only helpers in insurance or banking self-service That last one is the unlock for regulated industries. A few tools that read browser state, touch nothing, and either point the user to the right screen or pull the right info for their agent. But as a founder, you don't need any of that for your customers right away. Build it in-house first. Your internal tools, your own team, your own workflow. That's where the potential is.

The Startup Ideas Podcast (SIP) 🧃

10,592 views • 2 months ago

"Closed models are cheaper. The reason you need open models is that they give you control". - Nvidia CEO Jensen Huang "One of the things that people misunderstand about open models is this: yes, you can host them yourself, and you can build your own computer. But most people use computers in the cloud. Frankly, I think closed models are cheaper. If you do not build everything yourself, you avoid the significant cost and expertise required to train, fine-tune, maintain, guardrail, evaluate, and keep the model safe. You would also need to build or obtain the computers required to host it. So, there is nothing cheap about doing all of that yourself. The reason you need open models is that they give you control. You can adapt them for your own highly specialized use cases. I think there is a lot of misunderstanding about closed models versus open models. We felt it was important for people to understand that there is a place for both." ---- From "Bloomberg Podcasts" YouTube channel, (full video link in comment)

"Closed models are cheaper. The reason you need open models is that they give you control". - Nvidia CEO Jensen Huang "One of the things that people misunderstand about open models is this: yes, you can host them yourself, and you can build your own computer. But most people use computers in the cloud. Frankly, I think closed models are cheaper. If you do not build everything yourself, you avoid the significant cost and expertise required to train, fine-tune, maintain, guardrail, evaluate, and keep the model safe. You would also need to build or obtain the computers required to host it. So, there is nothing cheap about doing all of that yourself. The reason you need open models is that they give you control. You can adapt them for your own highly specialized use cases. I think there is a lot of misunderstanding about closed models versus open models. We felt it was important for people to understand that there is a place for both." ---- From "Bloomberg Podcasts" YouTube channel, (full video link in comment)

Rohan Paul

32,317 views • 2 months ago

"One of the biggest misconceptions" Cerebras CFO Bob Komin pushes back on the small-models narrative. "We serve all models, and there is no limit to the size of the models that we can serve. Today, we're serving trillion parameter models. We're serving trillion parameter models that are internal for OpenAI today. We are currently running OpenAI 5.4 and 5.5 with them."

"One of the biggest misconceptions" Cerebras CFO Bob Komin pushes back on the small-models narrative. "We serve all models, and there is no limit to the size of the models that we can serve. Today, we're serving trillion parameter models. We're serving trillion parameter models that are internal for OpenAI today. We are currently running OpenAI 5.4 and 5.5 with them."

Deirdre Bosa

84,608 views • 5 months ago

😼🔮💵📈 $GME “Now it's all about the transformation and that cash pile is growing, it seems to be growing, will they raise more money? We'll see, I don't know - I see it. But...if people, if you just change one factor and that's that you don't think Ryan Cohen is a doofus, alright? If you don't think he's a doofus and he's not a fool, if you don't think that - and all of a sudden, suddenly they're sitting on billions of dollars in cash, and you actually think not only is he not a doofus but - he's extremely talented in that he has the capability to turn this around, even if you don't agree with that, even if you don't believe it, that's fine, if you just pause for a moment and think about that - I think it really changes your mindset so that becomes the bet - that becomes the bet.”

😼🔮💵📈 $GME “Now it's all about the transformation and that cash pile is growing, it seems to be growing, will they raise more money? We'll see, I don't know - I see it. But...if people, if you just change one factor and that's that you don't think Ryan Cohen is a doofus, alright? If you don't think he's a doofus and he's not a fool, if you don't think that - and all of a sudden, suddenly they're sitting on billions of dollars in cash, and you actually think not only is he not a doofus but - he's extremely talented in that he has the capability to turn this around, even if you don't agree with that, even if you don't believe it, that's fine, if you just pause for a moment and think about that - I think it really changes your mindset so that becomes the bet - that becomes the bet.”

The GG Wizard Swami

32,029 views • 1 year ago

A James Wan horror film, directed in Agent One. Film + full breakdown, in the thread. Agent One is a unit of execution that works with the filmmaker. Don't miss the thread👇

A James Wan horror film, directed in Agent One. Film + full breakdown, in the thread. Agent One is a unit of execution that works with the filmmaker. Don't miss the thread👇

Invideo

248,573 views • 6 months ago

9 out of 10 multi-agent projects never leave demo mode Not because the model is bad. Because the structure is missing. Most people who try to build [ a team of AI agents ] end up with one agent talking to itself in five tabs. The agents don't share context → Don't divide work → Don't know what the others are doing. > stage 1: if your agent doesn't have a real loop, observe, act, iterate, you have a long prompt, not an agent > stage 2: subagents need isolated context, the orchestrator never reads their raw transcript, only the summary > stage 3: the orchestrator plans and delegates, the moment it executes, it drowns in details that belong inside subagents > stage 4: without a shared task list it's not a team, it's five agents duplicating each other's work in parallel > stage 5: a permissions file is what lets you sleep, the model cannot bypass it because the rule lives outside the model a team of AI agents is not more model, it is more structure the most ignored stage in every demo that died: - durability... when a 50-step task crashes at step 47 and starts from zero, that's not a model failure. that's a missing write-to-disk call

9 out of 10 multi-agent projects never leave demo mode Not because the model is bad. Because the structure is missing. Most people who try to build [ a team of AI agents ] end up with one agent talking to itself in five tabs. The agents don't share context → Don't divide work → Don't know what the others are doing. > stage 1: if your agent doesn't have a real loop, observe, act, iterate, you have a long prompt, not an agent > stage 2: subagents need isolated context, the orchestrator never reads their raw transcript, only the summary > stage 3: the orchestrator plans and delegates, the moment it executes, it drowns in details that belong inside subagents > stage 4: without a shared task list it's not a team, it's five agents duplicating each other's work in parallel > stage 5: a permissions file is what lets you sleep, the model cannot bypass it because the rule lives outside the model a team of AI agents is not more model, it is more structure the most ignored stage in every demo that died: - durability... when a 50-step task crashes at step 47 and starts from zero, that's not a model failure. that's a missing write-to-disk call

Shadow Nick

14,743 views • 5 months ago

.President Donald J. Trump: "See that plane? That's in perfect, mint condition... Everything should look that way. We work with the military; we do a real job. Look at the polish. Look at the way everything is beautiful — and that's the way we want D.C." DECLINE IS A CHOICE!

.President Donald J. Trump: "See that plane? That's in perfect, mint condition... Everything should look that way. We work with the military; we do a real job. Look at the polish. Look at the way everything is beautiful — and that's the way we want D.C." DECLINE IS A CHOICE!

Rapid Response 47

282,050 views • 5 months ago

Boris Cherny (Boris Cherny) on the Lightcone Podcast about how he approached building Claude Code: "At Anthropic, we don't build for the model of today, we build for the model of six months from now. And that's still my advice to founders that are building on LLMs."

Boris Cherny (Boris Cherny) on the Lightcone Podcast about how he approached building Claude Code: "At Anthropic, we don't build for the model of today, we build for the model of six months from now. And that's still my advice to founders that are building on LLMs."

Y Combinator

81,264 views • 8 months ago

Your agents suck when using the web because 85% of it doesn't have an API. Browserbase gives them everything they need to do work online. Leading AI companies like Ramp, Lovable, and Clay trust us to power agents that do real work on behalf of real people. With a single API key, your agent gets everything it needs to navigate the wild web: browsers, search, fetch, identity, a sandbox runtime, and model gateway. Stop waiting on integrations, build agents that can browse and interact with the web just like humans.

Your agents suck when using the web because 85% of it doesn't have an API. Browserbase gives them everything they need to do work online. Leading AI companies like Ramp, Lovable, and Clay trust us to power agents that do real work on behalf of real people. With a single API key, your agent gets everything it needs to navigate the wild web: browsers, search, fetch, identity, a sandbox runtime, and model gateway. Stop waiting on integrations, build agents that can browse and interact with the web just like humans.

Paul Klein IV

1,732,963 views • 7 months ago

Break the routine. Turn on your mental faculties Imagination, Memory, Intuition, Reason, Perception, and Will. Know that everything in nature does that except the human population because of the corporatization of everything. Don't let your education get in the way of your learning.

Break the routine. Turn on your mental faculties Imagination, Memory, Intuition, Reason, Perception, and Will. Know that everything in nature does that except the human population because of the corporatization of everything. Don't let your education get in the way of your learning.

Jason Mauck

12,447 views • 2 years ago

The worst feature of most products is one of our best features: customer support. It's incredible how much better a product becomes when you have an incredibly helpful, friendly, experienced, and deeply knowledgeable team on-call, for you, whenever you need them. Huge bonus. We don't outsource support. We don't see it as a cost center. Outstanding support is an asset, not a liability. A number of our support folks have been with us for over a decade! That's longer than most tech companies have even been in business! And the same team that works on Basecamp works on HEY. It's all the same A-team. We WANT you to contact us. We WANT to talk to you. Get in touch. You'll see the difference.

The worst feature of most products is one of our best features: customer support. It's incredible how much better a product becomes when you have an incredibly helpful, friendly, experienced, and deeply knowledgeable team on-call, for you, whenever you need them. Huge bonus. We don't outsource support. We don't see it as a cost center. Outstanding support is an asset, not a liability. A number of our support folks have been with us for over a decade! That's longer than most tech companies have even been in business! And the same team that works on Basecamp works on HEY. It's all the same A-team. We WANT you to contact us. We WANT to talk to you. Get in touch. You'll see the difference.

Jason Fried

301,767 views • 1 year ago

Are humanoids the largest addressable market in history? Andrew Kang says "to understand that you have to look at what is the total market for human labor. It's something like 50% of GDP, some people put it at $40 trillion or $60 trillion." "The way that I would think about it is a humanoid can do everything that a human can do in the future, but it doesn't need to rest, it doesn't need to take breaks, it doesn't need to go on vacation. It's not going to quit on you, so you don't need to hire and retrain somebody else."

Are humanoids the largest addressable market in history? Andrew Kang says "to understand that you have to look at what is the total market for human labor. It's something like 50% of GDP, some people put it at $40 trillion or $60 trillion." "The way that I would think about it is a humanoid can do everything that a human can do in the future, but it doesn't need to rest, it doesn't need to take breaks, it doesn't need to go on vacation. It's not going to quit on you, so you don't need to hire and retrain somebody else."

Anthony Pompliano 🌪

109,720 views • 5 months ago

🔧 Bring Your Own Model — and Let Agent Forge Handle the Rest! John plugged in his own custom model — and in minutes, it was up and running as a fully deployed AI Agent. With Agent Forge, you’re not limited to prebuilt tools. You bring the intelligence. We automate everything else. Real models. Real-time deployment. Zero-code friction. 🔗 Try it here:

🔧 Bring Your Own Model — and Let Agent Forge Handle the Rest! John plugged in his own custom model — and in minutes, it was up and running as a fully deployed AI Agent. With Agent Forge, you’re not limited to prebuilt tools. You bring the intelligence. We automate everything else. Real models. Real-time deployment. Zero-code friction. 🔗 Try it here:

AITECH CLOUD NETWORK

32,798 views • 1 year ago

AppleScript is the oldest way to drive a Mac. Frontier models butcher it. So we trained two small ones that don't. Open weights. On-device. 100% compile.

AppleScript is the oldest way to drive a Mac. Frontier models butcher it. So we trained two small ones that don't. Open weights. On-device. 100% compile.

Osaurus

54,111 views • 4 months ago

🚨 AWESOME! PRESIDENT TRUMP: "Take a look at that plane. See that plane? That's in perfect mint condition!" "Take a look at that. Look at that. Isn't that amazing? And EVERYTHING should look that way." 🔥🔥 "We work with the military. We do a real job. If you look at the polish, everything is beautiful." "And that's the way we want DC." 🇺🇸

🚨 AWESOME! PRESIDENT TRUMP: "Take a look at that plane. See that plane? That's in perfect mint condition!" "Take a look at that. Look at that. Isn't that amazing? And EVERYTHING should look that way." 🔥🔥 "We work with the military. We do a real job. If you look at the polish, everything is beautiful." "And that's the way we want DC." 🇺🇸

Eric Daugherty

47,917 views • 5 months ago

Once you don't focus on stuff to close the gaps of insecurity in your soul, you stop acting in a manner that doesn't match your moral compass and principles … This loop is everything and I hope you realize impressing those who you don't even like is the seed that leads to actions that you don't admire in yourself.

Once you don't focus on stuff to close the gaps of insecurity in your soul, you stop acting in a manner that doesn't match your moral compass and principles … This loop is everything and I hope you realize impressing those who you don't even like is the seed that leads to actions that you don't admire in yourself.

Gary Vaynerchuk

18,914 views • 8 months ago