고정된 트윗
Raytar
311 posts
Raytar
@Raytar
build with AI daily. wins, breaks, occasional bullshit. (the banner isn't a joke. building anyway.)
가입일 Eylül 2024
63 팔로잉5.7K 팔로워
My old classmate got called a "f*cking loser" from the street while he was sending a $500 invoice.
He found out from the video three days later. Laughed. Sent another invoice.
He builds websites in Google AI Studio, cold-calls businesses that don't know what AI is, and closes them at $500 plus $297 a month.
Last Tuesday he made $1,782 before lunch.
He's 22. No degree. No team. No code.
One AI tool. Six prompts. $20 monthly costs.
When I asked if he feels like he's missing out he just said "missing out on what?"
His full setup is in the article below. Not just the prompts. Everything.
Noisy@noisyb0y1
English
@Thihoanggs Not sure if it’s true, but Japanese accounts that reply right when a post drops always look like bots.
English
My ex's younger brother showed me his Shopify and it ruined my week.
+$278. +$124. +$68.
That was in three seconds.
He's fifteen.
He sells an AI-generated cookbook for $54.50 a pop. Last week: 1,080 orders, up 120%.
He was choosing colors for a 530-horsepower BMW while it ticked. He doesn't have a license.
Alpine White. The free color. On a hundred-grand car.
I asked about college.
He said "Grinding at 15, free at 17."
The four-hour video takes a new Shopify store to $50k a month. Link in the quoted post
Raytar@Raytar
He turned a new Shopify store into $50,000 in 30 days. Then recorded 4 hours showing the entire system. Every supplier, every angle, every tool. People sell this as a $497 course. He's giving it away for free below. Save this.
English
you use AI every day. you still cannot explain how it works.
1 free hour from Stanford fixes that.
after this hour you stop guessing. you know why ChatGPT lies. why it rambles. why it sounds so sure when it is wrong.
and your teacher is the guy who trains the models behind ChatGPT.
1 hour. free. it is right here.
people sell worse than this for $500.
save it for tonight. share it with someone who builds.
Raytar@Raytar
he tested 5760 architectures at Google for a full year. the winner was the original Transformer from 2017. Hyung Won Chung told that story at MIT with a small smile. then went to OpenAI and trained o1. 1 hour. free. by one of the few people on earth who actually moves the frontier. meanwhile your feed is full of guys writing architecture threads who have never trained a model anyone uses. he just told MIT that 99% of AI research is theater. your AI worldview was built by men who read his papers. badly. now you can read him directly. you will rewatch this. save it now.
English
Andrew Ng just told a room full of YC founders to stop hiring engineers.
co-founded Coursera. built Google Brain. ships a new AI startup every month. 35 so far.
Silicon Valley lived by 1 PM to 7 engineers for a decade. yesterday his team pitched 1 PM to 0.5 engineers. twice as many product people as engineers. he's never seen that before. nobody has.
his 35 startups already build this way.
the founders in that room already watched it.
you haven't.
he dropped his $190 million fund's entire playbook. on camera. for free.
1 hour. still up. your competitors already found it.
Raytar@Raytar
English
you're still prompting Claude like it's 2023.
"think step by step." it's trained into the model. saying it again wastes tokens on an instruction Claude already follows.
few-shot examples? Anthropic's own team says they make frontier models worse. you hand Claude a script when it already knows more than you do.
Jeremy Hadfield and Hannah Moran. Anthropic Applied AI. built Claude Code and Claude's research.
that prompting guide you bookmarked two years ago is now your biggest handicap.
you will rewatch this.
Raytar@Raytar
The full Claude course in 1 hour. Free. Anthropic pays Software Engineers $565K a year to ship on this stack: prompting, Claude Code, agent building, MCP integrations. Bookmark this. Watch it before your next Claude session. You'll know more about Claude than most teams figure out in a year.
English
he started coding at NASA JPL at 17. spent 7 years building dev tools at a Series C startup.
then tried every agent framework. LangChain. CrewAI. LangGraph. Griptape. threw them all away.
Dex Horthy coined "context engineering" in April 2025. by September, Anthropic published a full engineering guide on it. Karpathy adopted it. LangChain cited his repo.
19,000+ stars on GitHub. front page of Hacker News all day.
1 hour. free. by the person the biggest names in AI are now reading.
you will rewatch this.
Raytar@Raytar
"I was definitely the first prompt engineer at Anthropic. Might have been the first in the world." Alex Albert just spent 35 minutes explaining how they train Claude's personality from the inside. 35 minutes. free. by the person who invented the role. most people think Claude's character is a system prompt. it's not. you'll never look at Claude the same way.
English
the $1,000,000 answer to AI agents is a .md file.
not a framework. not a new model. a folder with a markdown file and some scripts.
Barry Zhang and Mahesh Murag. they created MCP and Agent Skills at Anthropic.
the analogy they used: you don't want a 300 IQ genius doing your taxes. you want the experienced tax professional.
agents today are the genius. brilliant but clueless. skills make them the expert.
lecture below.
16 minutes. free. replaces 10-hour $500 AI agent courses.
watch it tonight.
Raytar@Raytar
The full Claude course in 1 hour. Free. Anthropic pays Software Engineers $565K a year to ship on this stack: prompting, Claude Code, agent building, MCP integrations. Bookmark this. Watch it before your next Claude session. You'll know more about Claude than most teams figure out in a year.
English
Anthropic's CEO just explained on Lex Fridman how Claude actually works.
4 hours. free. with the CEO, the person who built Claude's character, and the founder of mechanistic interpretability.
most people use Claude every day at maybe 10% of what it can actually do. this video is the other 90%.
Save it.
This will change the way you use Claude forever.
Raytar@Raytar
The full Claude course in 1 hour. Free. Anthropic pays Software Engineers $565K a year to ship on this stack: prompting, Claude Code, agent building, MCP integrations. Bookmark this. Watch it before your next Claude session. You'll know more about Claude than most teams figure out in a year.
English
Anthropic team just spent 30 minutes explaining how they actually build agents with Claude internally.
22 minutes. free. by Alex Albert, Brad Abrams and Katelyn Lesse, who run the Claude Developer Platform.
most people build agents by stacking prompts and hoping. these three show you the patterns that survive contact with production.
Bookmark it & give it 30 minutes today.
Raytar@Raytar
The full Claude course in 1 hour. Free. Anthropic pays Software Engineers $565K a year to ship on this stack: prompting, Claude Code, agent building, MCP integrations. Bookmark this. Watch it before your next Claude session. You'll know more about Claude than most teams figure out in a year.
English
in 1929 sound came to film and 80% of silent film stars disappeared in eighteen months. they did not see it coming.
Google's head of AI walked into MIT and said we are exactly in 1929 again. except this time the silent film stars are us.
Doug Eck. Google DeepMind. 1 hour. free.
he gives you one sentence that explains every AI product that has ever launched. why some change the world and others die in beta. once you have it, you cannot look at the field the same way again.
then he shows you where you are standing on the curve. and the only choice you actually have left.
watch it before someone you work with does.
save this. you will rewatch it.
Raytar@Raytar
"I was definitely the first prompt engineer at Anthropic. Might have been the first in the world." Alex Albert just spent 35 minutes explaining how they train Claude's personality from the inside. 35 minutes. free. by the person who invented the role. most people think Claude's character is a system prompt. it's not. you'll never look at Claude the same way.
English
Anthropic's "Prompt Doctor" just spent 90 minutes rewriting real audience prompts live on camera.
90 minutes. free. by Zack Witten, who works on prompting at Anthropic.
most people prompt Claude like they're chatting with a friend. this is how the people who actually train it do it.
all in one video and completely free.
Raytar@Raytar
The full Claude course in 1 hour. Free. Anthropic pays Software Engineers $565K a year to ship on this stack: prompting, Claude Code, agent building, MCP integrations. Bookmark this. Watch it before your next Claude session. You'll know more about Claude than most teams figure out in a year.
English
two OpenAI researchers walked into Stanford and made every AI thread on your feed look like fan fiction.
Jason Wei invented Chain of Thought. Hyung Won Chung built o1.
1 hour. free.
Jason shows the chart that explains why AI suddenly gets new abilities. even OpenAI cannot predict when.
Hyung Won shows the one trend that controls the entire field. every AI headline fits one frame after this.
every other AI take on your feed is grown men playing dress up.
skip this and every AI thread feels hollow.
save this. you will rewatch it.
Raytar@Raytar
"I was definitely the first prompt engineer at Anthropic. Might have been the first in the world." Alex Albert just spent 35 minutes explaining how they train Claude's personality from the inside. 35 minutes. free. by the person who invented the role. most people think Claude's character is a system prompt. it's not. you'll never look at Claude the same way.
English
a Google researcher walked into MIT and made an AI do math correctly by adding seven words to the prompt.
the seven words: "you are an MIT mathematician."
drop them, model gets it wrong. add them, right. same model. same question. every time.
Carter Smith. runs Gemini at Google. 1 hour. free.
he then spent the next 50 minutes explaining why. it is the cleanest hour on how LLMs actually work I have seen in two years.
you will come back to this. save it now.
Raytar@Raytar
"I was definitely the first prompt engineer at Anthropic. Might have been the first in the world." Alex Albert just spent 35 minutes explaining how they train Claude's personality from the inside. 35 minutes. free. by the person who invented the role. most people think Claude's character is a system prompt. it's not. you'll never look at Claude the same way.
English
The OpenAI founding member just dropped a 40-minute talk on how software is fundamentally changing in 2026.
40 minutes. free. at YC Startup School. by Andrej Karpathy.
the talk that gave "vibe coding" its name. and the cleanest framing of where Claude, Cursor and AI coding are going that exists right now.
if you build with AI and skipped this, you're working off last year's mental model.
save it.
Raytar@Raytar
The full Claude course in 1 hour. Free. Anthropic pays Software Engineers $565K a year to ship on this stack: prompting, Claude Code, agent building, MCP integrations. Bookmark this. Watch it before your next Claude session. You'll know more about Claude than most teams figure out in a year.
English
he tested 5760 architectures at Google for a full year. the winner was the original Transformer from 2017.
Hyung Won Chung told that story at MIT with a small smile. then went to OpenAI and trained o1.
1 hour. free. by one of the few people on earth who actually moves the frontier.
meanwhile your feed is full of guys writing architecture threads who have never trained a model anyone uses. he just told MIT that 99% of AI research is theater.
your AI worldview was built by men who read his papers. badly. now you can read him directly.
you will rewatch this. save it now.
Raytar@Raytar
"I was definitely the first prompt engineer at Anthropic. Might have been the first in the world." Alex Albert just spent 35 minutes explaining how they train Claude's personality from the inside. 35 minutes. free. by the person who invented the role. most people think Claude's character is a system prompt. it's not. you'll never look at Claude the same way.
English
The best performing OnlyFans page I've seen this year isn't a person. It's four files and a $20 Claude subscription.
Some kid built her in Claude Code. Face, content, posting schedule, DMs. All automated. She doesn't exist.
$43,000 in 30 days. No camera. No model. No team.
He runs the whole thing from his bedroom. LED lights, a laptop, and Claude. That's it.
I spent 20 minutes looking at the page trying to figure out what was off. Nothing was off. That's the problem.
I texted him and asked for the files. He sent me everything. It's all in the article.
Raytar@Raytar
English
Anthropic researchers just spent 2h20min explaining how Claude actually thinks.
2h20min. free. by Sholto Douglas and Trenton Bricken, the two researchers who literally train Claude.
most people prompt Claude and hope. these two show you the mechanism behind why your prompts work or fail.
all in one video and completely free.
Raytar@Raytar
The full Claude course in 1 hour. Free. Anthropic pays Software Engineers $565K a year to ship on this stack: prompting, Claude Code, agent building, MCP integrations. Bookmark this. Watch it before your next Claude session. You'll know more about Claude than most teams figure out in a year.
English