Pinned Tweet
This is f*cking insane. Anthropic engineer Andrej Karpathy:
"The biggest mistake in AI right now is forcing agents to work before you master the harness around the model
We made that mistake at OpenAI in 2016. It cost us 5 years"
In 6 minutes, he explains why the agent is not the product. The foundation is
Opus 5.5 → Harness → Loops → Graphs → Agents That Emerge on Their Own
Demos are easy. Products take a decade
Worth more than most $1,000 agent courses
Bookmark and watch it today
Then read the "Harness, Loops, Graphs" guide below - build the foundation first ↓
fomoebka retweeted
This is f*cking insane. Anthropic engineer Andrej Karpathy:
"The biggest mistake in AI right now is forcing agents to work before you master the harness around the model
We made that mistake at OpenAI in 2016. It cost us 5 years"
In 6 minutes, he explains why the agent is not the product. The foundation is
Opus 5.5 → Harness → Loops → Graphs → Agents That Emerge on Their Own
Demos are easy. Products take a decade
Worth more than most $1,000 agent courses
Bookmark and watch it today
Then read the "Harness, Loops, Graphs" guide below - build the foundation first ↓
This is f*cking insane. Andrej Karpathy (ex-OpenAI, ex-Tesla AI lead):
"Agents aren't magic. They're distillation at scale
99.99% of your model's capacity is wasted on data it never needed"
In a 16-minute conversation, he lays out the stack that actually works: a small model, the right tools and a closed loop
Big Model → Small Model → Right Tools → Closed Loop → Self-Improving Agent
Most people keep paying for a bigger model. The loop is what makes a small one dangerous
Worth more than any $500 AI course this year
Bookmark and watch it today
Then read the "Harness, Loops, Graphs" guide below - the closed loop is the harness ↓
This is f*cking insane. Anthropic engineer Andrej Karpathy:
"The biggest mistake in AI right now is forcing agents to work before you master the harness around the model
We made that mistake at OpenAI in 2016. It cost us 5 years"
In 6 minutes, he explains why the agent is not the product. The foundation is
Opus 5.5 → Harness → Loops → Graphs → Agents That Emerge on Their Own
Demos are easy. Products take a decade
Worth more than most $1,000 agent courses
Bookmark and watch it today
Then read the "Harness, Loops, Graphs" guide below - build the foundation first ↓
This is f*cking insane. The head of Claude Code at Anthropic:
"85% of our engineers are running dozens or hundreds of agents
The way you do it is graph engineering"
One engineer at Anthropic now does the work of a whole team
In this 36-minute talk, he shows how far it's gone inside Anthropic and where it goes next
One Agent → Dozens of Agents → Graphs → Hundreds of Agents → Self-Improving System
Worth more than most $1,000 agent engineering courses
Bookmark and watch it today
Then read the "Harness, Loops, Graphs" guide below - graphs are the last layer ↓
This is f*cking insane. Anthropic engineer Andrej Karpathy:
"The biggest mistake in AI right now is forcing agents to work before you master the harness around the model
We made that mistake at OpenAI in 2016. It cost us 5 years"
In 6 minutes, he explains why the agent is not the product. The foundation is
Opus 5.5 → Harness → Loops → Graphs → Agents That Emerge on Their Own
Demos are easy. Products take a decade
Worth more than most $1,000 agent courses
Bookmark and watch it today
Then read the "Harness, Loops, Graphs" guide below - build the foundation first ↓
fomoebka retweeted
This is f*cking insane. The head of Claude Code at Anthropic:
"85% of our engineers are running dozens or hundreds of agents
The way you do it is graph engineering"
One engineer at Anthropic now does the work of a whole team
In this 36-minute talk, he shows how far it's gone inside Anthropic and where it goes next
One Agent → Dozens of Agents → Graphs → Hundreds of Agents → Self-Improving System
Worth more than most $1,000 agent engineering courses
Bookmark and watch it today
Then read the "Harness, Loops, Graphs" guide below - graphs are the last layer ↓
fomoebka retweeted
This is f*cking insane. The creator of Claude Code:
"Right now you still need to know how to code. In a year or two, it won't matter
I haven't edited a single line by hand since November"
The tool he built now writes 4% of every public commit on GitHub
In this 17-minute clip, Boris Cherny shows the exact setup he runs instead of writing code
Better than any $1,000 coding bootcamp I've seen
Save this. In a year we'll know if he was right
Then read the "Harness, Loops, Graphs" guide below ↓
This is f*cking insane. Anthropic Research Lead:
"Prompting is basically over
99% of our engineers already run swarms of 300+ self-improving agents. Now everyone is moving to graphs"
In just 20 minutes, she shows how graph engineering turns isolated agents into a system that improves itself
One Agent → Loops → Plan Mode → Graphs → Self-Improving System
Prompting was the old workflow. The graph around the model is the new one
Worth more than most $1,000 agent engineering courses
Bookmark and watch it today
Then read the "Harness, Loops, Graphs" guide below - graphs are the last layer ↓
This is f*cking insane. The creator of Claude Code:
"Right now you still need to know how to code. In a year or two, it won't matter
I haven't edited a single line by hand since November"
The tool he built now writes 4% of every public commit on GitHub
In this 17-minute clip, Boris Cherny shows the exact setup he runs instead of writing code
Better than any $1,000 coding bootcamp I've seen
Save this. In a year we'll know if he was right
Then read the "Harness, Loops, Graphs" guide below ↓
This is f*cking insane. Anthropic engineer:
"Prompting is basically over
95% of our engineers already run self-improving loops. The next step is graphs"
In just 10 minutes, she builds her whole Claude Code setup live, starting from an empty terminal
Empty Terminal → Agents → Loops → Graphs → Self-Improving System
Prompting was the old workflow. Engineering the system around the model is the new one
Worth more than most $1,000 agent engineering courses
Bookmark and watch it today
Then read the "Harness, Loops, Graphs" guide below and see which layer you're on ↓
This is f*cking insane. The creator of Claude Code:
"Right now you still need to know how to code. In a year or two, it won't matter
I haven't edited a single line by hand since November"
The tool he built now writes 4% of every public commit on GitHub
In this 17-minute clip, Boris Cherny shows the exact setup he runs instead of writing code
Better than any $1,000 coding bootcamp I've seen
Save this. In a year we'll know if he was right
Then read the "Harness, Loops, Graphs" guide below ↓
This is f*cking insane. The creator of Claude Code:
"Right now you still need to know how to code. In a year or two, it won't matter
I haven't edited a single line by hand since November"
The tool he built now writes 4% of every public commit on GitHub
In this 17-minute clip, Boris Cherny shows the exact setup he runs instead of writing code
Better than any $1,000 coding bootcamp I've seen
Save this. In a year we'll know if he was right
Then read the "Harness, Loops, Graphs" guide below ↓
fomoebka retweeted
This is f*cking insane. Anthropic engineer:
"Prompting is basically over
You're supposed to build a system that prompts itself. 90% of our engineers already have one"
In this 30-minute session, she builds an agent that remembers, fixes its own mistakes and gets sharper with every run
Prompt → Memory → Loops → Self-Correction → Self-Improving System
Prompting was the old workflow. The system around the model is the new one
Worth more than most $1,000 agent engineering courses
Bookmark and watch it today
Then read the "Harness, Loops, Graphs" guide below - that's the system part ↓
This is f*cking insane. Anthropic engineer:
"Prompting is basically over
You're supposed to build a system that prompts itself. 90% of our engineers already have one"
In this 30-minute session, she builds an agent that remembers, fixes its own mistakes and gets sharper with every run
Prompt → Memory → Loops → Self-Correction → Self-Improving System
Prompting was the old workflow. The system around the model is the new one
Worth more than most $1,000 agent engineering courses
Bookmark and watch it today
Then read the "Harness, Loops, Graphs" guide below - that's the system part ↓
This is f*cking insane. Anthropic engineer:
"Prompting is basically over
You're supposed to build a system that prompts itself. 90% of our engineers already have one"
In this 30-minute session, she builds an agent that remembers, fixes its own mistakes and gets sharper with every run
Prompt → Memory → Loops → Self-Correction → Self-Improving System
Prompting was the old workflow. The system around the model is the new one
Worth more than most $1,000 agent engineering courses
Bookmark and watch it today
Then read the "Harness, Loops, Graphs" guide below - that's the system part ↓
This is f*cking insane. Anthropic engineer:
"Prompting is basically over
You're supposed to build a system that prompts itself. 90% of our engineers already have one"
In this 30-minute session, she builds an agent that remembers, fixes its own mistakes and gets sharper with every run
Prompt → Memory → Loops → Self-Correction → Self-Improving System
Prompting was the old workflow. The system around the model is the new one
Worth more than most $1,000 agent engineering courses
Bookmark and watch it today
Then read the "Harness, Loops, Graphs" guide below - that's the system part ↓
fomoebka retweeted
This is f*cking insane. Andrej Karpathy (ex-OpenAI, ex-Tesla AI lead):
"Agents aren't magic. They're distillation at scale
99.99% of your model's capacity is wasted on data it never needed"
In a 16-minute conversation, he lays out the stack that actually works: a small model, the right tools and a closed loop
Big Model → Small Model → Right Tools → Closed Loop → Self-Improving Agent
Most people keep paying for a bigger model. The loop is what makes a small one dangerous
Worth more than any $500 AI course this year
Bookmark and watch it today
Then read the "Harness, Loops, Graphs" guide below - the closed loop is the harness ↓
This is f*cking insane. Andrej Karpathy (ex-OpenAI, ex-Tesla AI lead):
"Agents aren't magic. They're distillation at scale
99.99% of your model's capacity is wasted on data it never needed"
In a 16-minute conversation, he lays out the stack that actually works: a small model, the right tools and a closed loop
Big Model → Small Model → Right Tools → Closed Loop → Self-Improving Agent
Most people keep paying for a bigger model. The loop is what makes a small one dangerous
Worth more than any $500 AI course this year
Bookmark and watch it today
Then read the "Harness, Loops, Graphs" guide below - the closed loop is the harness ↓
must to watch ↓
teapaw.org/fomoebka/status/211511…
This is f*cking insane. Creator of Claude Code:
"I barely write prompts anymore
I write loops that keep Claude working until the tests say it's actually done"
In a 1-hour podcast, Boris breaks down how Anthropic engineers run dozens of agents without babysitting a single one
Prompt → Harness → Loops → Routines → Self-Improving Agents
A loop doesn't make the model smarter. It makes every mistake cheap, visible and fixed before you ever see it
A routine starts the loop on its own, so the work keeps going while you sleep
Most people run 100 agents and get 100 confident wrong answers. The 1% add the check first
Worth more than most $1,500 Claude Code courses
Bookmark and watch it today
Then read the full Harness, Loops and Graphs guide below ↓
fomoebka retweeted
This is f*cking insane. Anthropic Research Lead:
"Prompting is basically over
99% of our engineers already run swarms of 300+ self-improving agents. Now everyone is moving to graphs"
In just 20 minutes, she shows how graph engineering turns isolated agents into a system that improves itself
One Agent → Loops → Plan Mode → Graphs → Self-Improving System
Prompting was the old workflow. The graph around the model is the new one
Worth more than most $1,000 agent engineering courses
Bookmark and watch it today
Then read the "Harness, Loops, Graphs" guide below - graphs are the last layer ↓
must to watch ↓
teapaw.org/fomoebka/status/211583…
This is f*cking insane. Anthropic Research Lead:
"Prompting is basically over
99% of our engineers already run swarms of 300+ self-improving agents. Now everyone is moving to graphs"
In just 20 minutes, she shows how graph engineering turns isolated agents into a system that improves itself
One Agent → Loops → Plan Mode → Graphs → Self-Improving System
Prompting was the old workflow. The graph around the model is the new one
Worth more than most $1,000 agent engineering courses
Bookmark and watch it today
Then read the "Harness, Loops, Graphs" guide below - graphs are the last layer ↓