AI, agents and emerging tech sharing what I learn, test and discover member @CELE8EX
Kazakhstan
Joined February 2024
- Tweets1.1K
- Following197
- Followers759
- Likes2K
Dmitro Cashback retweeted
This is f*cking insane. Chamath just exposed where AI revenue really comes from:
"1% of companies represent 90% of AI spending
And of that 90%, 80% of them are from unprofitable VC backed companies"
In 2 minutes, he walks through exactly how the AI money loop works:
0:36 - The 1% / 90% / 80% math
0:48 - The full circular cycle
1:21 - Why it's a structural problem
1:51 - What's different from early AWS
LPs → VCs → startups → AI tokens → "revenue" → same VCs
"75 cents of every dollar is actually just a startup sending it back and forth so that they can mark themselves up"
When AI revenue charts go vertical, most people see demand
@chamath sees a bunch of money losing entities spinning money around
AWS had Amazon retail behind it. This cycle doesn't
Better than any $1,000 AI investing course I've seen
Is AI demand real or just VC money on a loop?
This is f*cking insane. Chamath just exposed where AI revenue really comes from:
"1% of companies represent 90% of AI spending
And of that 90%, 80% of them are from unprofitable VC backed companies"
In 2 minutes, he walks through exactly how the AI money loop works:
0:36 - The 1% / 90% / 80% math
0:48 - The full circular cycle
1:21 - Why it's a structural problem
1:51 - What's different from early AWS
LPs → VCs → startups → AI tokens → "revenue" → same VCs
"75 cents of every dollar is actually just a startup sending it back and forth so that they can mark themselves up"
When AI revenue charts go vertical, most people see demand
@chamath sees a bunch of money losing entities spinning money around
AWS had Amazon retail behind it. This cycle doesn't
Better than any $1,000 AI investing course I've seen
Is AI demand real or just VC money on a loop?
Dmitro Cashback retweeted
AI firms need to raise 30% of what the Treasury will issue through '27
AI vs the US government for the same dollars. Who wins?
Chamath: The AI Boom Has a Borrowing Problem
“The entire United States economy is levered to the AI complex.
So very narrowly, we're talking about seven or eight companies that set the tone for the entire country. And whether we like it or not, that is where we are.
Why do I say that? The most important companies in the world are in a massive spending binge, right?
These were highly profitable, highly cash accretive, highly cash flow generative businesses that have said, ‘We're going to put the foot down on spending because we believe in what's coming into the future.’ That's what drives GDP and it's what drives growth.
Now, as they do that though, they deplete their cash and they have to raise debt.
The AI complex now owes $340-400 billion to the market. And you'd say, ‘Okay, why is that important?’
The problem is this: These companies are raising so much money it starts to crowd out the capital markets.
And the most important part of the capital market that it starts to crowd out, is the appetite for government treasuries.
And what you can see here is that when you look at these AI firms, the amount of money that they need to raise through the end of '27, it's 30% of all of what the Treasury's about to issue in that same period.
So, whether we like it or not, they are going to start competing for demand.”
AI firms need to raise 30% of what the Treasury will issue through '27
AI vs the US government for the same dollars. Who wins?
Chamath: The AI Boom Has a Borrowing Problem
“The entire United States economy is levered to the AI complex.
So very narrowly, we're talking about seven or eight companies that set the tone for the entire country. And whether we like it or not, that is where we are.
Why do I say that? The most important companies in the world are in a massive spending binge, right?
These were highly profitable, highly cash accretive, highly cash flow generative businesses that have said, ‘We're going to put the foot down on spending because we believe in what's coming into the future.’ That's what drives GDP and it's what drives growth.
Now, as they do that though, they deplete their cash and they have to raise debt.
The AI complex now owes $340-400 billion to the market. And you'd say, ‘Okay, why is that important?’
The problem is this: These companies are raising so much money it starts to crowd out the capital markets.
And the most important part of the capital market that it starts to crowd out, is the appetite for government treasuries.
And what you can see here is that when you look at these AI firms, the amount of money that they need to raise through the end of '27, it's 30% of all of what the Treasury's about to issue in that same period.
So, whether we like it or not, they are going to start competing for demand.”
Dmitro Cashback retweeted
MARC ANDREESSEN: AI MIGHT BE A THIRD KIND OF CAPITALISM
He breaks business history into two phases
→ phase 1: "name on the door". Henry Ford runs Ford and "just tells everybody what to do"
→ the problem: "Henry Ford doesn't scale"
→ phase 2: professional managers, "trained to be experts in management", not in the product
Managerialism "basically took over everything". Big companies, governments, nonprofits
VC was the protest movement against it:
"we're like raging against the dying of the light"
Now AI opens a third option:
founder's spark of genius + AI superpowers for "all the managerial stuff"
Because what are bots best at?
"filling out forms… writing reports… doing all the managerial work. Like they're amazing at it"
His verdict for the incumbents: "figure out innovation or die trying"
Are managers the first job AI really replaces?
MARC ANDREESSEN: AI MIGHT BE A THIRD KIND OF CAPITALISM
He breaks business history into two phases
→ phase 1: "name on the door". Henry Ford runs Ford and "just tells everybody what to do"
→ the problem: "Henry Ford doesn't scale"
→ phase 2: professional managers, "trained to be experts in management", not in the product
Managerialism "basically took over everything". Big companies, governments, nonprofits
VC was the protest movement against it:
"we're like raging against the dying of the light"
Now AI opens a third option:
founder's spark of genius + AI superpowers for "all the managerial stuff"
Because what are bots best at?
"filling out forms… writing reports… doing all the managerial work. Like they're amazing at it"
His verdict for the incumbents: "figure out innovation or die trying"
Are managers the first job AI really replaces?
Dmitro Cashback retweeted
IS THIS ALL JUST A CIRCLE JERK?
Jon Stewart asked it. Ed Zitron, host of Better Offline, walked him through the money behind AI
"OpenAI and Anthropic are really interesting companies, economically speaking, because they lose billions of dollars"
His version of the chain:
→ AI labs lose billions, so banks get uneasy lending to them
→ so they rent compute instead of building it
→ Anthropic will be paying "$3.1 billion a month" to Elon Musk's SpaceX as of December, he says
→ Oracle is building "about $400 billion worth of AI data centers for one customer"
And NVIDIA sits in the middle of everything:
"They invest in all the companies"
CoreWeave "takes out debt using the chips as collateral and then buys more chips using that"
Lab → rents compute → cloud borrows on chips → buys more chips → NVIDIA invests in the lab
"It's magical"
Bubble or the biggest buildout in history?
IS THIS ALL JUST A CIRCLE JERK?
Jon Stewart asked it. Ed Zitron, host of Better Offline, walked him through the money behind AI
"OpenAI and Anthropic are really interesting companies, economically speaking, because they lose billions of dollars"
His version of the chain:
→ AI labs lose billions, so banks get uneasy lending to them
→ so they rent compute instead of building it
→ Anthropic will be paying "$3.1 billion a month" to Elon Musk's SpaceX as of December, he says
→ Oracle is building "about $400 billion worth of AI data centers for one customer"
And NVIDIA sits in the middle of everything:
"They invest in all the companies"
CoreWeave "takes out debt using the chips as collateral and then buys more chips using that"
Lab → rents compute → cloud borrows on chips → buys more chips → NVIDIA invests in the lab
"It's magical"
Bubble or the biggest buildout in history?
Dmitro Cashback retweeted
your AI ads are giving it away
left is what chatgpt makes. right is the same headphones in krea ad studio, from one store link
which one are you buying from? first few ads are free, link below
🤝 Paid partnership
Dmitro Cashback retweeted
DARIO SAW IT COMING IN 2019
Anthropic CEO Dario Amodei on the first time he saw GPT-2:
"you just scale this thing up another factor of a million and it'll be better than humans at everything"
Most people shrugged. It could translate a bit, summarize a doc, write a few lines of code
→ "maybe 10% of the people around me had exactly the same reaction"
→ "and it was often the smartest ones"
→ the rest dismissed it
His rule for every new model since:
hard today → easy in 6 months → impossible becomes routine in 12
The people who called it pseudoscience back then?
"all of them are applying for jobs at the LLM companies"
"we remember what you were saying in 2019. You have no credibility on this matter"
The lesson isn't about GPT-2. It's that the limits you see today are the worst the models will ever be
What are you still dismissing right now?
DARIO SAW IT COMING IN 2019
Anthropic CEO Dario Amodei on the first time he saw GPT-2:
"you just scale this thing up another factor of a million and it'll be better than humans at everything"
Most people shrugged. It could translate a bit, summarize a doc, write a few lines of code
→ "maybe 10% of the people around me had exactly the same reaction"
→ "and it was often the smartest ones"
→ the rest dismissed it
His rule for every new model since:
hard today → easy in 6 months → impossible becomes routine in 12
The people who called it pseudoscience back then?
"all of them are applying for jobs at the LLM companies"
"we remember what you were saying in 2019. You have no credibility on this matter"
The lesson isn't about GPT-2. It's that the limits you see today are the worst the models will ever be
What are you still dismissing right now?
Dmitro Cashback retweeted
I spent $17 one night finding out which model should animate a single car.
$2 went to Kimi K3.1 Pro, $15 to Opus 5.5. The $2 render is the one I kept.
That's most of my feed: same prompt, a few tabs, receipts. Except the receipts lived in my head and in a pile of screenshots.
So I hired someone. Viktor is an AI employee, and his job is my build log.
Day one, I handed him the pile. What came back in Slack:
-> 4 builds logged: the model, the cost, and whether I kept the result.
-> $31.50 spent across them, $28 of it on renders I threw away.
-> The model that earned its money: Kimi K3.1 Pro.
That's $28 I stop spending next month.
He runs on the best model available each month, so he isn't loyal to any of them. Neither am I.
He proposes, I decide. Nothing gets posted or paid for without my yes.
Hire @viktor_com for your team. Start free.
#AIemployee #AItools
Paid Partnership
🤝 Paid partnership
Dmitro Cashback retweeted
ELON JUST SAID THE QUIET PART
Elon Musk with Walter Isaacson:
"I find it unlikely that humans would remain in control if there's a million times more digital intelligence than there is biological intelligence"
"So I hope the AI is nice to us"
His fix isn't a kill switch, it's the training data:
→ AIs are trained on everything online
→ "they read 80 versions of the Terminator script"
→ so write more good futures than dystopias
Doom stories → training data → doom behavior
He wants "protopia": not a perfect future, just a good one you'd recognize and say yes to
If models learn from what we write, every dystopian thread is a vote
Are we writing the Terminator into existence?
ELON JUST SAID THE QUIET PART
Elon Musk with Walter Isaacson:
"I find it unlikely that humans would remain in control if there's a million times more digital intelligence than there is biological intelligence"
"So I hope the AI is nice to us"
His fix isn't a kill switch, it's the training data:
→ AIs are trained on everything online
→ "they read 80 versions of the Terminator script"
→ so write more good futures than dystopias
Doom stories → training data → doom behavior
He wants "protopia": not a perfect future, just a good one you'd recognize and say yes to
If models learn from what we write, every dystopian thread is a vote
Are we writing the Terminator into existence?
Dmitro Cashback retweeted
This is a f*cking goldmine..
An Anthropic engineer:
"90% of our engineers already run self-improving loops on Opus 5.5. Now everyone is moving to agentic graphs
Prompting is basically done".
In 10 minutes she builds her entire Claude Code setup live, starting from an empty terminal.
Agents -> Loops -> Graphs -> Self-Improving Systems
Prompting was the old way to work.
Graph engineering is the next one.
This 10-minute breakdown beats most $700 agent engineering courses.
Watch it today.
Then read the article below ↓
Dmitro Cashback retweeted
Andrej Karpathy:
"Vibe coding is amazing. But Agentic + Opus 5.5 harness engineering is the next level.
99% of my coding routine is automated by AI agents."
In this 30-minute talk, Andrej Karpathy explains how to build an AI agent workflow from scratch.
Worth more than any $500 agentic engineering course on the internet.
Watch it today, then read The Karpathy Playbook below ↓
Dmitro Cashback retweeted
Elon showing up when someone builds a full RPG with no studio, no team and no publisher
I SPENT 3 BILLION TOKENS AND 1 MONTH BUILDING THIS GAME, WITH ZERO GAMEDEV EXPERIENCE
So what came out of it?
EXTRA: Ascension - a single-player story RPG I'm building with Claude Opus 5.5
> a world with its own night sky and a Great Tree on the horizon
> an orc hero who fights with two axes
> creatures that Claude built from just 3 images each
No studio, no team, no publisher
This is month one, I'm building the rest in public, follow along
Caelora fell. The world ascended.
Coming to PC in 2027
Dmitro Cashback retweeted
Andy Jassy on the best deal Amazon ever made
"It's one of the very best acquisitions we've done in the history of the company, if not the best"
The Amazon CEO on Annapurna Labs, in an a16z interview:
→ Amazon didn't buy Annapurna for AI chips, it bought a team that was great at building chips
→ that team built Graviton, used by 99% of Amazon's top 1,000 EC2 customers
→ Graviton is about 40% more price-performant than other x86 processors
→ a separate Annapurna team then built Trainium, about 30 to 40% more price-performant than comparable processors
Buy a great team → let it build the next chip → own your compute
The lesson isn't the chip. It's that the team was the asset
Which acquisition will look this smart in 10 years?
Andy Jassy on the best deal Amazon ever made
"It's one of the very best acquisitions we've done in the history of the company, if not the best"
The Amazon CEO on Annapurna Labs, in an a16z interview:
→ Amazon didn't buy Annapurna for AI chips, it bought a team that was great at building chips
→ that team built Graviton, used by 99% of Amazon's top 1,000 EC2 customers
→ Graviton is about 40% more price-performant than other x86 processors
→ a separate Annapurna team then built Trainium, about 30 to 40% more price-performant than comparable processors
Buy a great team → let it build the next chip → own your compute
The lesson isn't the chip. It's that the team was the asset
Which acquisition will look this smart in 10 years?
Dmitro Cashback retweeted
Blackstone's president on why AI is not a bubble
"In a classic investment bubble, what you get is enormous spend of capital to build capacity to meet some estimated future demand"
Jon Gray's case, from Blackstone's Market Views:
→ railroads in the 19th century and telecom in the 20th were built ahead of demand, with a lot of leverage
→ with AI it's the exact opposite: demand is exploding and supply is behind
→ their companies are spending 20 times more on models, and individual usage keeps going up
→ supply is stuck on permits, community pushback, turbines, power plants, chips and raising tens of billions
Bubble: supply ahead of demand. AI: demand ahead of supply
The scarce thing today, he says, is AI infrastructure
Bubble or bottleneck: which one are we in?