This is f*cking insane. Fields Medal winner Terence Tao, the greatest living mathematician:
"We've learned almost nothing mathematically from what they did. It's like strip mining the landscape. You get all the fruits, but then the landscape is barren afterwards"
In under 2 minutes, he kills the "AI solved it, so we're winning" story: a 100-page proof that no human fully understands yet, bought with millions of dollars of compute
Output → Proof → Understanding → Harness → Work You Can Actually Trust
An answer nobody can explain is a result you can't build on
A harness is what makes the model show its work, check it and leave you something you understand
Everyone is cheering the answers. Almost nobody builds the system that makes them usable
Bookmark it, watch it tonight, then read the harness engineering guide below ↓
Your agent passed every test
The issue is still open
A better prompt won't fix it, because the model isn't the broken part
I broke down the harness, loop and graph around it: what each one owns, how they nest, and which one to fix first ↓ https://t.co/rcmwM1jrfB
This is f*cking insane. Fields Medal winner Terence Tao, the greatest living mathematician:
"We've learned almost nothing mathematically from what they did. It's like strip mining the landscape. You get all the fruits, but then the landscape is barren afterwards"
In under 2 minutes, he kills the "AI solved it, so we're winning" story: a 100-page proof that no human fully understands yet, bought with millions of dollars of compute
Output → Proof → Understanding → Harness → Work You Can Actually Trust
An answer nobody can explain is a result you can't build on
A harness is what makes the model show its work, check it and leave you something you understand
Everyone is cheering the answers. Almost nobody builds the system that makes them usable
Bookmark it, watch it tonight, then read the harness engineering guide below ↓
This is f*cking insane. Jeff Dean, Google's chief scientist and the engineer behind TensorFlow:
"Some people have this illusion that you're going to be able to run this loop infinitely fast. I just find that kind of implausible
It's not going to make them go infinitely fast"
In 22 seconds, he kills the "AI rewrites itself into superintelligence overnight" story everyone keeps posting
The intelligence explosion isn't coming to save you. The speed-up comes from the loop you build yourself
Model → Agents → Harness → Checks → Loops → A System That Gets Better Every Week
The model won't improve itself overnight. Your harness can improve it every single day
Everyone is waiting for the explosion. The few building loops are already compounding
These 22 seconds are worth more than most $1,000 AI courses
Bookmark and watch it today
Then read the harness engineering guide below ↓
Your agent passed every test
The issue is still open
A better prompt won't fix it, because the model isn't the broken part
I broke down the harness, loop and graph around it: what each one owns, how they nest, and which one to fix first ↓ https://t.co/rcmwM1jrfB
This is f*cking insane. Jeff Dean, Google's chief scientist and the engineer behind TensorFlow:
"Some people have this illusion that you're going to be able to run this loop infinitely fast. I just find that kind of implausible
It's not going to make them go infinitely fast"
In 22 seconds, he kills the "AI rewrites itself into superintelligence overnight" story everyone keeps posting
The intelligence explosion isn't coming to save you. The speed-up comes from the loop you build yourself
Model → Agents → Harness → Checks → Loops → A System That Gets Better Every Week
The model won't improve itself overnight. Your harness can improve it every single day
Everyone is waiting for the explosion. The few building loops are already compounding
These 22 seconds are worth more than most $1,000 AI courses
Bookmark and watch it today
Then read the harness engineering guide below ↓
This is f*cking insane. Jeff Dean, Google's chief scientist and the engineer behind TensorFlow:
"Some people have this illusion that you're going to be able to run this loop infinitely fast. I just find that kind of implausible
It's not going to make them go infinitely fast"
In 22 seconds, he kills the "AI rewrites itself into superintelligence overnight" story everyone keeps posting
The intelligence explosion isn't coming to save you. The speed-up comes from the loop you build yourself
Model → Agents → Harness → Checks → Loops → A System That Gets Better Every Week
The model won't improve itself overnight. Your harness can improve it every single day
Everyone is waiting for the explosion. The few building loops are already compounding
These 22 seconds are worth more than most $1,000 AI courses
Bookmark and watch it today
Then read the harness engineering guide below ↓
This is f*cking insane. Nobel Prize winner Geoffrey Hinton, the godfather of AI:
"The idea that it's just predicting the next word doesn't make any sense. It had to understand what was going on"
In just 40 seconds, he kills the "it's just autocomplete" excuse with one experiment: a programming language the model had never seen, and it still got the answer right
Autocomplete → Understanding → Agents → Harness → Real Work Done
A model that understands can be handed a job, not just a prompt
A harness is what turns that understanding into work it finishes on its own
Anyone can call it autocomplete. Almost nobody builds the system that puts it to work
Watch it today, then save the full harness engineering guide below before everyone catches up ↓
Your hourly check finds nothing 23 hours a day
You still pay a top model for every one of those runs
A cheaper model won't fix it, a split will
I broke down how to set the mode once per channel with Viktor and script the steps that never needed AI ↓ https://t.co/zGbETTN0OI
This is f*cking insane. Nobel Prize winner Geoffrey Hinton, the godfather of AI:
"The idea that it's just predicting the next word doesn't make any sense. It had to understand what was going on"
In just 40 seconds, he kills the "it's just autocomplete" excuse with one experiment: a programming language the model had never seen, and it still got the answer right
Autocomplete → Understanding → Agents → Harness → Real Work Done
A model that understands can be handed a job, not just a prompt
A harness is what turns that understanding into work it finishes on its own
Anyone can call it autocomplete. Almost nobody builds the system that puts it to work
Watch it today, then save the full harness engineering guide below before everyone catches up ↓
This is f*cking insane. Andrej Karpathy, OpenAI co-founder and the man who ran AI at Tesla for 5 years:
"I don't see AI as a distinct technology. You can't find computers or mobile phones in GDP. Everything slowly diffuses
With AI, we're going to see the exact same thing. It's just more automation. AI is still fundamentally a program"
In 60 seconds, he kills the "AGI changes everything overnight" story: no jump, no singularity, just the same curve we have been on for decades
So nobody gets handed the future on launch day. It goes to the people who automate their own work first
Program → Automation → Agents → Harness → Work That Compounds
A program runs once. A harness runs it every day, checks every step and gets better each week
Everyone is waiting for the big AI moment. It already happened, it just looks boring
This one minute is worth more than most $1,000 AI courses
Bookmark and watch it today
Then read the harness engineering guide below ↓
This is f*cking insane. Andrej Karpathy, OpenAI co-founder and the man who ran AI at Tesla for 5 years:
"I don't see AI as a distinct technology. You can't find computers or mobile phones in GDP. Everything slowly diffuses
With AI, we're going to see the exact same thing. It's just more automation. AI is still fundamentally a program"
In 60 seconds, he kills the "AGI changes everything overnight" story: no jump, no singularity, just the same curve we have been on for decades
So nobody gets handed the future on launch day. It goes to the people who automate their own work first
Program → Automation → Agents → Harness → Work That Compounds
A program runs once. A harness runs it every day, checks every step and gets better each week
Everyone is waiting for the big AI moment. It already happened, it just looks boring
This one minute is worth more than most $1,000 AI courses
Bookmark and watch it today
Then read the harness engineering guide below ↓
Your agent passed every test
The issue is still open
A better prompt won't fix it, because the model isn't the broken part
I broke down the harness, loop and graph around it: what each one owns, how they nest, and which one to fix first ↓ https://t.co/rcmwM1jrfB
This is f*cking insane. Former OpenAI CTO Mira Murati just explained why every AI you use today is deaf and blind half the time:
"While they're thinking, it's almost like they're deaf and blind. They cannot perceive anything else about what's going on
Interaction models are not turn-based. They're continuously taking in audio, text, video and continuously providing output"
In 2 minutes, she shows the model Thinking Machines built to catch interruptions and silence in 20-millisecond chunks
Turn-Based Chat → Real-Time Models → Agents That Listen → Humans In The Loop
Most people still type a prompt and wait. The next wave of agents works next to you while you talk
This 2-minute clip is worth more than most $1,000 AI courses
Bookmark and watch it today
Then read the harness engineering guide below ↓
This is f*cking insane. Former OpenAI CTO Mira Murati just explained why every AI you use today is deaf and blind half the time:
"While they're thinking, it's almost like they're deaf and blind. They cannot perceive anything else about what's going on
Interaction models are not turn-based. They're continuously taking in audio, text, video and continuously providing output"
In 2 minutes, she shows the model Thinking Machines built to catch interruptions and silence in 20-millisecond chunks
Turn-Based Chat → Real-Time Models → Agents That Listen → Humans In The Loop
Most people still type a prompt and wait. The next wave of agents works next to you while you talk
This 2-minute clip is worth more than most $1,000 AI courses
Bookmark and watch it today
Then read the harness engineering guide below ↓
This is f*cking insane. Bill Gates just said what it will take for the world to finally take AI seriously:
"If you have a big cyberattack that caused a lot of problems, people will say, okay, they were right
It may be that we have to suffer a few of the negatives before we respond as fully as we should"
In just 60 seconds, he admits the risk of letting a few AI disasters happen first is "very high"
AI → Agents → Cyberattacks → Catastrophe → Rules
The first ones to get hurt will be the people running agents with nothing checking them
A harness decides what an agent can touch, checks every step, and stops it before it breaks something
Save and watch it today
Then save the harness engineering guide below before the first one hits ↓
This is f*cking insane. Bill Gates just said what it will take for the world to finally take AI seriously:
"If you have a big cyberattack that caused a lot of problems, people will say, okay, they were right
It may be that we have to suffer a few of the negatives before we respond as fully as we should"
In just 60 seconds, he admits the risk of letting a few AI disasters happen first is "very high"
AI → Agents → Cyberattacks → Catastrophe → Rules
The first ones to get hurt will be the people running agents with nothing checking them
A harness decides what an agent can touch, checks every step, and stops it before it breaks something
Save and watch it today
Then save the harness engineering guide below before the first one hits ↓
This is f*cking insane. Turing Award winner Yann LeCun, who ran AI research at Meta for 12 years:
"There's absolutely no way in hell. The idea that we're going to have a country of geniuses in a data center, that's complete BS
It would feel like you have a PhD sitting next to you. But it's not a PhD. It's a system with a gigantic memory and retrieval ability, not a system that can invent solutions to new problems"
In 60 seconds, he kills the "just scale it" plan: bigger LLMs alone are not getting anyone to human-level AI
So the edge was never waiting for a smarter model. It's the system you build around the one you already have
Memory → Retrieval → Agents → Harness → Loops → New Problems Solved
The model remembers. The harness is what makes it work
Everyone is waiting for AGI. The few building harnesses are already shipping with what exists today
This one minute is worth more than most $1,000 AI courses
Bookmark and watch it today
Then read the harness engineering guide below ↓
This is f*cking insane. Turing Award winner Yann LeCun, who ran AI research at Meta for 12 years:
"There's absolutely no way in hell. The idea that we're going to have a country of geniuses in a data center, that's complete BS
It would feel like you have a PhD sitting next to you. But it's not a PhD. It's a system with a gigantic memory and retrieval ability, not a system that can invent solutions to new problems"
In 60 seconds, he kills the "just scale it" plan: bigger LLMs alone are not getting anyone to human-level AI
So the edge was never waiting for a smarter model. It's the system you build around the one you already have
Memory → Retrieval → Agents → Harness → Loops → New Problems Solved
The model remembers. The harness is what makes it work
Everyone is waiting for AGI. The few building harnesses are already shipping with what exists today
This one minute is worth more than most $1,000 AI courses
Bookmark and watch it today
Then read the harness engineering guide below ↓
Holy sh*t, ex-Google CEO Eric Schmidt just said what nobody in AI says out loud:
"If you really want to make money, it's actually easy. Build an agent to do something
Build a better agent than anyone else and you can probably build one incredible company and own it yourself"
He calls it the agentic period: for the next year or two, everyone builds agents and they all compete
You don't get there by reading AI threads, you get there by learning agents, Claude Code, prompts, memory, skills, MCP and routines
Prompts → Claude Code → Memory → Skills → MCP → A Company Run By Agents
Bookmark this before your feed buries it
Every link below is free:
Agent architecture
https://t.co/UXmhZELboH
Claude Code 101
https://t.co/eG58P8BIIV
Claude Code in action
https://t.co/6rGzMp9r7S
Prompt engineering
https://t.co/EfXhkYcRS7
Interactive prompt course
https://t.co/Sjy7EJCJsB
Claude.md + memory
https://t.co/XEn7mLNNrd
Skills
https://t.co/leVrZrnYKk
MCP
https://t.co/rt0ZUx0GLt
Routines
https://t.co/RlWcxK7M2i
Claude Code ultimate guide
https://t.co/icuq37T4UD
Awesome Claude Code
https://t.co/chvNxwr2ac
Anthropic academy
https://t.co/WTvUKVR5fq
Official Claude Code docs
https://t.co/LVgc2lrxR1
All of it costs $0
Most people will keep asking AI one question at a time
Bookmark it and start on the first link tonight
Then see how the harness around an agent really works in the harness engineering guide below ↓
This is f*cking insane. SpaceXAI founder Elon Musk just told everyone who works at a screen what's coming:
"Anything that is digital, which is just someone at a computer doing something, AI is going to take over those jobs like lightning
Anything that's physically moving atoms, plumbing, welding, cooking, farming, those jobs will exist for a much longer time"
In just 5 minutes, he explains why he calls AI a supersonic tsunami
Desk Jobs → AI Agents → Autonomous Work → Working Becomes Optional
Coding, email, answering the phone, anything that's just a person at a screen goes first
Moving pixels is the work AI takes, moving atoms is the work that lasts
The only screen job that stays safe is the one running the agents
Worth more than any $2,000 AI course
Bookmark and watch it today
Then read how to be the one running the agents in the harness engineering guide below ↓
This is f*cking insane. SpaceXAI founder Elon Musk just told everyone who works at a screen what's coming:
"Anything that is digital, which is just someone at a computer doing something, AI is going to take over those jobs like lightning
Anything that's physically moving atoms, plumbing, welding, cooking, farming, those jobs will exist for a much longer time"
In just 5 minutes, he explains why he calls AI a supersonic tsunami
Desk Jobs → AI Agents → Autonomous Work → Working Becomes Optional
Coding, email, answering the phone, anything that's just a person at a screen goes first
Moving pixels is the work AI takes, moving atoms is the work that lasts
The only screen job that stays safe is the one running the agents
Worth more than any $2,000 AI course
Bookmark and watch it today
Then read how to be the one running the agents in the harness engineering guide below ↓