fomoebka (@fomoebka) header
fomoebka (@fomoebka) profile picture

ai // vibecoding // development

Joined July 2011
Photo by fomoebka (@fomoebka)Photo by fomoebka (@fomoebka)Photo by fomoebka (@fomoebka)Photo by fomoebka (@fomoebka)Photo by fomoebka (@fomoebka)Photo by fomoebka (@fomoebka)Photo by fomoebka (@fomoebka)Photo by fomoebka (@fomoebka)Photo by fomoebka (@fomoebka)Photo by fomoebka (@fomoebka)
> be Boris Cherny > write the O'Reilly book Programming TypeScript > spend years as an engineer at Meta > join Anthropic in 2024 > build a little terminal tool as a side project > it becomes Claude Code > half of Silicon Valley starts coding with it > leave for Cursor > come back to Anthropic two weeks later > now say "I don't prompt Claude anymore. My job is to create loops" he built the tool everyone uses, then stopped using it the way everyone does. give him 30 minutes, then read the full guide ↓
Your agent is not autonomous You are its loop You check every step because nothing else in the system can I broke down the 7 levels from babysitting agents to running a fleet ↓
6
41
1,204
> be Dario Amodei > grow up in San Francisco > study physics at Stanford > PhD in biophysics at Princeton > 2014: join Baidu to work on speech AI with Andrew Ng > then Google Brain > then OpenAI, VP of Research, help build GPT-2 and GPT-3 > walk out with his sister Daniela and a handful of colleagues > start Anthropic > build Claude > now taking it public he helped build GPT-3, then walked out to build Claude. give him 37 minutes on the smart way to use it, then read the full guide ↓
Your agent is not autonomous You are its loop You check every step because nothing else in the system can I broke down the 7 levels from babysitting agents to running a fleet ↓
2
1
3
56
15,517
This is f*cking insane. The engineer who built Claude Code made a 28-minute video on how to write prompts that actually work CLAUDE.md Files → Memory Shortcuts → Parallel Sessions → Prompting Patterns $300 courses don't cover what he shows in the first 10 minutes Codex users keep asking why people switch to Claude Code. This video is the answer Bookmark and watch it today Then read the next step after prompts: Harness, Loops, Graphs ↓
Your agent is not autonomous You are its loop You check every step because nothing else in the system can I broke down the 7 levels from babysitting agents to running a fleet ↓
1
55
3,919
This is f*cking insane. Anthropic went from 2,500 code changes a day to more than 30,000. Lydia Hallie from the Claude Code team explains what changed: "If you do want the guarantee, you can only enforce that in the harness" In 17 minutes, she shows how their engineers stopped approving every command and started running dozens of sessions for hours Shouting IMPORTANT in CLAUDE.md → Deny Rules → Sandbox → Hooks → Dozens of Sessions Without You Most of last year's prompting tricks now hold the models back Worth more than most $1,000 Claude Code courses Bookmark and watch it today Then read how to build that harness: Harness, Loops, Graphs ↓
Your agent is not autonomous You are its loop You check every step because nothing else in the system can I broke down the 7 levels from babysitting agents to running a fleet ↓
4
2
1
47
1,307
This is f*cking insane. Days before Anthropic goes public, its CEO Dario Amodei was asked what a kid finishing high school should do. No sugarcoating: "AI can do entry-level law work. It can do entry-level finance. It can do entry-level consulting It will also eliminate jobs, and probably a large number of them" In 42 seconds, he names the first careers in the line of fire His own number: half of entry-level white-collar jobs disrupted in 1 to 5 years Chatting → Prompting → Agents → Harness → You Run the AI Using AI is the bare minimum now. Running it is the skill Bookmark and watch it today Then read where to start: Harness, Loops, Graphs ↓
Your agent is not autonomous You are its loop You check every step because nothing else in the system can I broke down the 7 levels from babysitting agents to running a fleet ↓
1
33
718
This is f*cking insane. Anthropic CEO Dario Amodei sat down with JPMorgan CEO Jamie Dimon days before the IPO: "In 2023, models couldn't even do high school math. In 2025 they started solving unsolved problems in mathematics Next come the problems you win a Fields Medal for" In this 38-minute conversation, he covers why compute is his deepest business dilemma, why all seven cofounders are still there, and how the Fortune 100 already runs on Claude High School Math → Unsolved Problems → Fields Medal Problems → Solving Biology He told the biggest banker in America to underpromise and overdeliver. Then he said this Worth more than any $2,000 course on where AI is going Bookmark and watch it before the IPO Then read what people build on his models today: Harness, Loops, Graphs ↓
Your agent is not autonomous You are its loop You check every step because nothing else in the system can I broke down the 7 levels from babysitting agents to running a fleet ↓
4
1
40
1,503
This is f*cking insane. Anthropic engineers: "We stopped asking one agent to do everything One agent plans, one builds, one judges. They cycle until the app actually works" In 40 minutes, the team behind Claude Code builds a full app from scratch with that loop on Fable 5.2 Plan → Build → Judge → Retry → An App That Actually Works The winners won't have the smartest model. They'll have the best loop Worth more than any $1,500 agent course Bookmark and watch it today Then read the full guide: Harness, Loops, Graphs ↓
Your agent is not autonomous You are its loop You check every step because nothing else in the system can I broke down the 7 levels from babysitting agents to running a fleet ↓
39
2,155
This is f*cking insane. Anthropic CEO Dario Amodei, on the IPO roadshow: "Cancer, Alzheimer's, heart disease, mental illness. These will be relics of the past in less than a decade If we do this right, we can create an age of miracles" In 54 seconds, the man behind Claude and Fable 5.2 says problems that beat us for millennia get solved in under 10 years Chatbots → Agents → Harness → Loops → Decades of Research in a Few Years This is what he is telling investors right now, not a podcast guess The research he means is done by agents. You can run the same kind today Save this before it disappears from your feed Every repo below is free: An agent that researches any topic in depth github.com/assafelovic/gpt-r… Full research reports written by agents github.com/stanford-oval/sto… Anthropic's plugins for legal, finance, sales, HR and data work github.com/anthropics/knowle… Anthropic's agents for analyst work github.com/anthropics/financ… Skills for Excel, Word, PowerPoint and PDF github.com/anthropics/skills Claude Agent SDK github.com/anthropics/claude… A harness with everything built in github.com/langchain-ai/deep… 12 principles of agents that actually work github.com/humanlayer/12-fac… Agent memory github.com/thedotmack/claude… A loop that runs until the job is done github.com/snarktank/ralph Browser control github.com/microsoft/playwri… Claude cookbooks github.com/anthropics/claude… Free agents course, 18 lessons github.com/microsoft/ai-agen… Free Anthropic courses github.com/anthropics/course… All of this costs $0 Most people will scroll past it. The ones building agents today will be inside it Then read the full guide: Harness, Loops, Graphs ↓
Your agent is not autonomous You are its loop You check every step because nothing else in the system can I broke down the 7 levels from babysitting agents to running a fleet ↓
2
2
45
31,202
This is f*cking insane. Anthropic just put its full IPO pitch online. CEO Dario Amodei, to investors: "Cancer, Alzheimer's, heart disease, mental illness. These will be relics of the past in less than a decade If we do this right, we can create an age of miracles" In this 42-minute roadshow video, the company behind Claude and Fable 5.2 explains where $33B of revenue came from and what it builds next Chatbot → Claude Code → Agents → Compute → A Country of Geniuses He says "compute" 49 times. That is the whole bet Worth more than any $2,000 course on where AI is going Bookmark and watch it before the IPO on November 19 Then read what people already build on it: Harness, Loops, Graphs ↓
1
37
This is f*cking insane. Anthropic just put its full IPO pitch online. CEO Dario Amodei, to investors: "Cancer, Alzheimer's, heart disease, mental illness. These will be relics of the past in less than a decade If we do this right, we can create an age of miracles" In this 42-minute roadshow video, the company behind Claude and Fable 5.2 explains where $33B of revenue came from and what it builds next Chatbot → Claude Code → Agents → Compute → A Country of Geniuses He says "compute" 49 times. That is the whole bet Worth more than any $2,000 course on where AI is going Bookmark and watch it before the IPO on November 19 Then read what people already build on it: Harness, Loops, Graphs ↓
Your agent is not autonomous You are its loop You check every step because nothing else in the system can I broke down the 7 levels from babysitting agents to running a fleet ↓
4
4
1
52
4,080
This is f*cking insane. One of my friends makes $1.2M a year as an Anthropic engineer Fable 5.2 dropped yesterday, so I asked him how their team actually prompts it He sent me a video that was never meant to be public: the core team's prompting playbook "We don't write prompts for Fable. We write the system that writes them" 33 minutes. Halfway through I realized I'd been using Claude wrong for two years Prompt → Context → Harness → Loops → A System That Prompts Itself Worth more than most $2,000 agent engineering courses Bookmark and watch it today Then read the "Harness, Loops, Graphs" guide below ↓
Your agent is not autonomous You are its loop You check every step because nothing else in the system can I broke down the 7 levels from babysitting agents to running a fleet ↓
4
14
1
152
28,838
This is f*cking insane. Anthropic CEO Dario Amodei, on the IPO roadshow: "Cancer, Alzheimer's, heart disease, mental illness. These will be relics of the past in less than a decade If we do this right, we can create an age of miracles" In 54 seconds, the man behind Claude and Fable 5.2 says problems that beat us for millennia get solved in under 10 years Chatbots → Agents → Harness → Loops → Decades of Research in a Few Years This is what he is telling investors right now, not a podcast guess The research he means is done by agents. You can run the same kind today Save this before it disappears from your feed Every repo below is free: An agent that researches any topic in depth github.com/assafelovic/gpt-r… Full research reports written by agents github.com/stanford-oval/sto… Anthropic's plugins for legal, finance, sales, HR and data work github.com/anthropics/knowle… Anthropic's agents for analyst work github.com/anthropics/financ… Skills for Excel, Word, PowerPoint and PDF github.com/anthropics/skills Claude Agent SDK github.com/anthropics/claude… A harness with everything built in github.com/langchain-ai/deep… 12 principles of agents that actually work github.com/humanlayer/12-fac… Agent memory github.com/thedotmack/claude… A loop that runs until the job is done github.com/snarktank/ralph Browser control github.com/microsoft/playwri… Claude cookbooks github.com/anthropics/claude… Free agents course, 18 lessons github.com/microsoft/ai-agen… Free Anthropic courses github.com/anthropics/course… All of this costs $0 Most people will scroll past it. The ones building agents today will be inside it Then read the full guide: Harness, Loops, Graphs ↓
1
108
This is f*cking insane. Anthropic Research Lead: "We stopped prompting Claude months ago Every engineer on our team runs 500+ self-improving agents wired into one graph. The graph fixes itself overnight" In 19 minutes, she shows how a graph turns a pile of separate agents into one system that gets better every run One Agent → Loops → Plan Mode → Graphs → Self-Improving System A prompt dies after one answer. A graph keeps every lesson Worth more than most $2,000 agent engineering courses Bookmark and watch it today Then read the "Harness, Loops, Graphs" guide below, the graph is the last layer ↓
Your agent is not autonomous You are its loop You check every step because nothing else in the system can I broke down the 7 levels from babysitting agents to running a fleet ↓
2
5
1
104
16,740
This is f*cking insane. Anthropic CEO Dario Amodei, on the IPO roadshow: "Cancer, Alzheimer's, heart disease, mental illness. These will be relics of the past in less than a decade If we do this right, we can create an age of miracles" In 54 seconds, the man behind Claude and Fable 5.2 says problems that beat us for millennia get solved in under 10 years Chatbots → Agents → Harness → Loops → Decades of Research in a Few Years This is what he is telling investors right now, not a podcast guess The research he means is done by agents. You can run the same kind today Save this before it disappears from your feed Every repo below is free: An agent that researches any topic in depth github.com/assafelovic/gpt-r… Full research reports written by agents github.com/stanford-oval/sto… Anthropic's plugins for legal, finance, sales, HR and data work github.com/anthropics/knowle… Anthropic's agents for analyst work github.com/anthropics/financ… Skills for Excel, Word, PowerPoint and PDF github.com/anthropics/skills Claude Agent SDK github.com/anthropics/claude… A harness with everything built in github.com/langchain-ai/deep… 12 principles of agents that actually work github.com/humanlayer/12-fac… Agent memory github.com/thedotmack/claude… A loop that runs until the job is done github.com/snarktank/ralph Browser control github.com/microsoft/playwri… Claude cookbooks github.com/anthropics/claude… Free agents course, 18 lessons github.com/microsoft/ai-agen… Free Anthropic courses github.com/anthropics/course… All of this costs $0 Most people will scroll past it. The ones building agents today will be inside it Then read the full guide: Harness, Loops, Graphs ↓
24
fomoebka retweeted
This is f*cking insane. Anthropic Research Lead: "We stopped prompting Claude months ago Every engineer on our team runs 500+ self-improving agents wired into one graph. The graph fixes itself overnight" In 19 minutes, she shows how a graph turns a pile of separate agents into one system that gets better every run One Agent → Loops → Plan Mode → Graphs → Self-Improving System A prompt dies after one answer. A graph keeps every lesson Worth more than most $2,000 agent engineering courses Bookmark and watch it today Then read the "Harness, Loops, Graphs" guide below, the graph is the last layer ↓
Your agent is not autonomous You are its loop You check every step because nothing else in the system can I broke down the 7 levels from babysitting agents to running a fleet ↓
2
5
1
104
16,740
fomoebka retweeted
fable 5.2 after replacing opus 5.5
Introducing Claude Fable 5.2 and Claude Mythos 5.2, designed to tackle the most daunting technical challenges. They’re capable collaborators on ambitious projects in many fields, and show improvements in areas like coding, reasoning, and vision.
13
2
2
80
7,762
This is f*cking insane. Anthropic CEO Dario Amodei was asked what an 18-year-old should do right now: "AI can already do entry-level law, entry-level finance, entry-level consulting. It will eliminate a lot of those jobs" His timeline: half of entry-level white-collar work disrupted in 1 to 5 years In under 2 minutes, he gives the one move that still works School → Internship → Entry-Level Job → Done By AI → Run The AI Or Report To Someone Who Does The kids who learn to run agents now will hire the ones who didn't Worth more than a $60,000 year of college Bookmark it and send it to someone graduating this year Then read the skill to learn first: the 7 levels from babysitting agents to running a fleet ↓
Your agent is not autonomous You are its loop You check every step because nothing else in the system can I broke down the 7 levels from babysitting agents to running a fleet ↓
5
11
3
92
25,987
This is f*cking insane. Anthropic Research Lead: "We stopped prompting Claude months ago Every engineer on our team runs 500+ self-improving agents wired into one graph. The graph fixes itself overnight" In 19 minutes, she shows how a graph turns a pile of separate agents into one system that gets better every run One Agent → Loops → Plan Mode → Graphs → Self-Improving System A prompt dies after one answer. A graph keeps every lesson Worth more than most $2,000 agent engineering courses Bookmark and watch it today Then read the "Harness, Loops, Graphs" guide below, the graph is the last layer ↓
1
1
92
fomoebka retweeted
This is f*cking insane. Anthropic CEO Dario Amodei was asked what an 18-year-old should do right now: "AI can already do entry-level law, entry-level finance, entry-level consulting. It will eliminate a lot of those jobs" His timeline: half of entry-level white-collar work disrupted in 1 to 5 years In under 2 minutes, he gives the one move that still works School → Internship → Entry-Level Job → Done By AI → Run The AI Or Report To Someone Who Does The kids who learn to run agents now will hire the ones who didn't Worth more than a $60,000 year of college Bookmark it and send it to someone graduating this year Then read the skill to learn first: the 7 levels from babysitting agents to running a fleet ↓
Your agent is not autonomous You are its loop You check every step because nothing else in the system can I broke down the 7 levels from babysitting agents to running a fleet ↓
5
11
3
92
25,987
This is f*cking insane. Anthropic engineer: "You're not supposed to prompt Opus 5.5. You're supposed to build the system that prompts it for you" In 33 minutes, she shows how Anthropic builds agents that remember, catch their own mistakes and get sharper every run One Prompt → Harness → Memory → Self-Check Loop → An Agent That Improves Itself Everyone is still polishing prompts. The people who build the model stopped doing that a year ago Worth more than any $2,000 agent course Bookmark and watch it today Then read the motion engineering breakdown: "Your AI video is not bad. Your checks are." ↓
I fixed the same bug in two films Opus never remembers my fixes So I hired an AI employee to remember them for me, then tried to make him break my one rule on deadline He didn't ↓
6
14
167
31,278
This is f*cking insane. Anthropic CEO Dario Amodei was asked what an 18-year-old should do right now: "AI can already do entry-level law, entry-level finance, entry-level consulting. It will eliminate a lot of those jobs" His timeline: half of entry-level white-collar work disrupted in 1 to 5 years In under 2 minutes, he gives the one move that still works School → Internship → Entry-Level Job → Done By AI → Run The AI Or Report To Someone Who Does The kids who learn to run agents now will hire the ones who didn't Worth more than a $60,000 year of college Bookmark it and send it to someone graduating this year Then read the skill to learn first: the 7 levels from babysitting agents to running a fleet ↓
15
This is f*cking insane. Anthropic engineer: "Prompting is basically over. 95% of our engineers already run self-improving loops, and the next move is graphs" In 10 minutes, she builds her whole Claude Code setup live, starting from an empty terminal Empty Terminal → Agents → Loops → Graphs → A System That Improves Itself The people building Opus 5.5 stopped writing prompts. They engineer the system around it Worth more than most $1,500 agent engineering courses Bookmark and watch it today Then read the motion engineering breakdown: "Your AI video is not bad. Your checks are." ↓
I fixed the same bug in two films Opus never remembers my fixes So I hired an AI employee to remember them for me, then tried to make him break my one rule on deadline He didn't ↓
4
3
1
47
2,604
This is f*cking insane. Anthropic CEO Dario Amodei was asked what an 18-year-old should do right now: "AI can already do entry-level law, entry-level finance, entry-level consulting. It will eliminate a lot of those jobs" His timeline: half of entry-level white-collar work disrupted in 1 to 5 years In under 2 minutes, he gives the one move that still works School → Internship → Entry-Level Job → Done By AI → Run The AI Or Report To Someone Who Does The kids who learn to run agents now will hire the ones who didn't Worth more than a $60,000 year of college Bookmark it and send it to someone graduating this year Then read the skill to learn first: the 7 levels from babysitting agents to running a fleet ↓
64