wtf4uk (@wtf4uk) header
wtf4uk (@wtf4uk) profile picture

23 ✞ ˚₊‧꒰ა ai researcher & builder ‧₊˚ ♡⃕ ⭒ dm always open

Joined April 2023
Photo by wtf4uk (@wtf4uk)Photo by wtf4uk (@wtf4uk)Photo by wtf4uk (@wtf4uk)Photo by wtf4uk (@wtf4uk)Photo by wtf4uk (@wtf4uk)Photo by wtf4uk (@wtf4uk)Photo by wtf4uk (@wtf4uk)Photo by wtf4uk (@wtf4uk)Photo by wtf4uk (@wtf4uk)Photo by wtf4uk (@wtf4uk)
Pinned Tweet
This is f*cking insane. Microsoft AI CEO Mustafa Suleyman just put a date on the end of office work: "Most of those tasks will be fully automated by an AI within the next 12 to 18 months" He means lawyers, accountants, project managers, marketers. Anyone who works at a computer In 92 seconds, he also says models that take $100,000, start a business and turn it into a million are coming this year Doing the Work → Checking the Work → Directing Agents → Harness → Running the Whole Job Save this before it disappears from your feed Everything below is free: Building effective agents anthropic.com/engineering/bu… Effective harnesses for long-running agents anthropic.com/engineering/ef… Effective context engineering for AI agents anthropic.com/engineering/ef… Writing tools for agents anthropic.com/engineering/wr… Free Anthropic courses github.com/anthropics/course… Total cost: $0 AI won't take your desk. Someone running a harness will 95% will wait for the 18 months to pass, the 5% who build a harness now will be the ones directing the work Bookmark it and send it to anyone who works at a computer Then use the guide below to build your harness before the deadline ↓
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
40
634
This is f*cking insane. Andrej Karpathy: "Vibe coding was fun. Harness engineering with Opus 5.5 is the next level I barely touch the keyboard now. 99% of my coding runs on agents" In 30 minutes, he builds an agent workflow from zero and shows where most people's setup breaks Vibe Coding → One Agent → Harness → Loops → Agents That Code While You Sleep The model isn't the bottleneck anymore. Your harness is Worth more than any $1,000 agentic engineering course Bookmark and watch it today Then read how to build that harness layer by layer in 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 ↓
5
1
53
1,418
This is f*cking insane. Anthropic engineer Addy Osmani (ex-Google, $500K-$1.2M a year): "Fable 5.2 doesn't make your code safe. The system around it does 24 skills, and the agent can't ship until it proves the work" In 41 minutes, he builds that system from scratch: idea, spec, small tasks, test-first, then the agent reviews itself in a browser Vague Idea → Spec → Small Tasks → Test-First Build → Self-Review → Shipped Without You He open-sourced all 24 skills, and the repo is close to 100K stars Worth more than most $2,000 agent engineering courses Bookmark and watch it today Then read how to build that harness layer by layer: Harness, Loops, Graphs ↓
14
This is f*cking insane. Anthropic engineer Addy Osmani (ex-Google, $500K-$1.2M a year): "Fable 5.2 doesn't make your code safe. The system around it does 24 skills, and the agent can't ship until it proves the work" In 41 minutes, he builds that system from scratch: idea, spec, small tasks, test-first, then the agent reviews itself in a browser Vague Idea → Spec → Small Tasks → Test-First Build → Self-Review → Shipped Without You He open-sourced all 24 skills, and the repo is close to 100K stars Worth more than most $2,000 agent engineering courses Bookmark and watch it today Then read how to build that harness layer by layer: 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
1
42
526
wtf4uk 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
78
7,321
This is f*cking insane. Andrej Karpathy: "The biggest mistake in AI right now is forcing agents before you've built the harness around the model We did exactly that at OpenAI in 2016 and lost 5 years" In under 6 minutes, he explains why an agent demo takes a weekend and an agent product takes a decade Prompt → Agent → Harness → Loops → Agents That Actually Ship Nobody buys the agent. They buy the harness that keeps it from breaking Worth more than most $1,500 agent courses Bookmark and watch it today Then read the harness that turns this lecture into a system: 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
1
44
623
wtf4uk retweeted
This is f*cking insane. X just dropped its own official Grok Bot guide for creators: "We use Grok Bot internally to produce the ads we boost across our social channels" In 6 minutes, it walks through the exact setup X's creator team runs, from the first idea to a posted video Brain Dump → Idea Breakdowns → Script Coach → Full Edit → Posted Everywhere When the platform itself shows you how it's done, that's worth one evening Worth more than a $1,500 content creation bootcamp Bookmark it and set it up tonight Then read the motion engineering article: "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 ↓
2
1
43
552
This is f*cking insane. X just dropped its own official Grok Bot guide for creators: "We use Grok Bot internally to produce the ads we boost across our social channels" In 6 minutes, it walks through the exact setup X's creator team runs, from the first idea to a posted video Brain Dump → Idea Breakdowns → Script Coach → Full Edit → Posted Everywhere When the platform itself shows you how it's done, that's worth one evening Worth more than a $1,500 content creation bootcamp Bookmark it and set it up tonight Then read the motion engineering article: "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 ↓
2
1
43
552
wtf4uk retweeted
This is f*cking insane. Demis Hassabis spent 16 years building DeepMind and won a Nobel Prize for it: "In 5 years, one person who really understands AI will outbuild an entire startup team" In this 1-hour talk, he lays out who comes out ahead in what's coming, and who doesn't Chatbots → Agents → Workflows → Automated Departments → One-Person Companies You probably don't have an hour right now. That's exactly why most people won't watch it People pay $15,000 for bootcamps that teach less than this Bookmark it and watch it tonight Then read "Your AI video is not bad. Your checks are.", the QC layer of motion engineering with Opus 5.5 ↓
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 ↓
3
1
44
563
This is f*cking insane. Demis Hassabis spent 16 years building DeepMind and won a Nobel Prize for it: "In 5 years, one person who really understands AI will outbuild an entire startup team" In this 1-hour talk, he lays out who comes out ahead in what's coming, and who doesn't Chatbots → Agents → Workflows → Automated Departments → One-Person Companies You probably don't have an hour right now. That's exactly why most people won't watch it People pay $15,000 for bootcamps that teach less than this Bookmark it and watch it tonight Then read "Your AI video is not bad. Your checks are.", the QC layer of motion engineering with Opus 5.5 ↓
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 ↓
3
1
44
563
wtf4uk retweeted
This is f*cking insane. Marc Andreessen, the man who built Netscape and a16z: "Founders are going to have superpowers they've simply never had before If I'd had these tools at Netscape in 1994... oh my God" He says entire industries are up for grabs, and the AI-native startups see it first The tools are free. Start here: Y Combinator's startup library ycombinator.com/library Startup School startupschool.org/ Do things that don't scale paulgraham.com/ds.html How to get startup ideas paulgraham.com/startupideas.… What YC wants to fund right now ycombinator.com/rfs Build the product fast: Claude Code 101 anthropic.skilljar.com/claud… Building effective agents anthropic.com/research/build… All of it costs $0 Bookmark it and build something this weekend 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 ↓
3
2
49
861
This is f*cking insane. Marc Andreessen, the man who built Netscape and a16z: "Founders are going to have superpowers they've simply never had before If I'd had these tools at Netscape in 1994... oh my God" He says entire industries are up for grabs, and the AI-native startups see it first The tools are free. Start here: Y Combinator's startup library ycombinator.com/library Startup School startupschool.org/ Do things that don't scale paulgraham.com/ds.html How to get startup ideas paulgraham.com/startupideas.… What YC wants to fund right now ycombinator.com/rfs Build the product fast: Claude Code 101 anthropic.skilljar.com/claud… Building effective agents anthropic.com/research/build… All of it costs $0 Bookmark it and build something this weekend 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 ↓
3
2
49
861
This is f*cking insane. Google Brain founder Andrew Ng: "Most people stop after the first agent and call it automation The last 40 minutes are the graph that runs it without you" In 2 hours, he goes from your first agent to a full graph system: 9:14 - Your first agent 33:11 - Loop engineering 1:02:46 - Graph engineering 1:30:15 - Agents that rewrite themselves 1:49:05 - The full graph system Same model, same tokens, a completely different week Worth more than every agent tutorial you bookmarked this year 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 ↓
47
897
wtf4uk retweeted
This is f*cking insane. Anthropic engineer: "90% of our engineers were running self-improving loops. Now everyone has moved to graphs Nobody here prompts anymore" In just 10 minutes, she builds her whole Claude Code setup live, from a blank terminal Blank Terminal → Agents → Loops → Graphs → Self-Improving System Prompting was the old workflow. The graph is the new one Worth more than most $1,000 agent 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 ↓
2
1
54
1,435
This is f*cking insane. Anthropic engineer: "90% of our engineers were running self-improving loops. Now everyone has moved to graphs Nobody here prompts anymore" In just 10 minutes, she builds her whole Claude Code setup live, from a blank terminal Blank Terminal → Agents → Loops → Graphs → Self-Improving System Prompting was the old workflow. The graph is the new one Worth more than most $1,000 agent 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 ↓
2
1
54
1,435
wtf4uk retweeted
Anthropic just showed a 28-minute workshop on how to actually prompt Claude Taught by the people who built it. Free. No sign-up. No paywall I've seen $300 courses that don't cover what they teach in the first 8 minutes Weak Prompt → Clear Task → The Right Context → A Prompt That Works Every Time Bookmark it and watch it today Then read the "Harness, Loops, Graphs" guide below - then turn that prompt into a loop ↓
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
44
604
Anthropic just showed a 28-minute workshop on how to actually prompt Claude Taught by the people who built it. Free. No sign-up. No paywall I've seen $300 courses that don't cover what they teach in the first 8 minutes Weak Prompt → Clear Task → The Right Context → A Prompt That Works Every Time Bookmark it and watch it today Then read the "Harness, Loops, Graphs" guide below - then turn that prompt into a loop ↓
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
44
604
wtf4uk retweeted
This is f*cking insane. Google Brain founder Andrew Ng: "100% of my tasks are done by AI agents. Self-improving loops are next Give it 3-6 months and prompting is gone" The man who taught machine learning to millions just stopped prompting In 31 minutes, he shows how to build a self-improving agent from scratch Better than any $500 agent course I've seen Bookmark and watch it today Then read the "Harness, Loops, Graphs" guide below - that's the loops part ↓
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
41
811
This is f*cking insane. Google Brain founder Andrew Ng: "100% of my tasks are done by AI agents. Self-improving loops are next Give it 3-6 months and prompting is gone" The man who taught machine learning to millions just stopped prompting In 31 minutes, he shows how to build a self-improving agent from scratch Better than any $500 agent course I've seen Bookmark and watch it today Then read the "Harness, Loops, Graphs" guide below - that's the loops part ↓
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
41
811
wtf4uk retweeted
This is f*cking insane. Anthropic CEO Dario Amodei, the man who left OpenAI to build Claude: "50% of entry-level white-collar jobs will be gone by 2030 Our lead Claude Code engineer hasn't written a single line of code in 2 months" 90% of Anthropic's own engineers already ship code with Claude In this 25-minute talk, he explains what's coming in the next 1-2 years and who comes out ahead Chatbot → Agents → Self-Improving Loops → Engineers Who Never Touch Code The jobs don't vanish for the people who run the agents Worth more than any $1,500 AI career course Bookmark and watch it today Then read the "Harness, Loops, Graphs" guide below - that's how you become the one running them ↓
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
37
824