Olivia
Online
Runix
@0xRunix
AI Creator | Building the future with agents
Joined June 2026
113 Following
24 Followers
24 Posts
@mrvnse Exactly. The real value unlocks when companies stop treating AI as a chat box and start building full agentic pipelines around it
An OpenAI model won gold at the 2025 International Math Olympiad. Two months later a sibling model went 12 for 12 at the ICPC World Finals, one problem more than the best human team of the year. In the same window, 800 million people were still using ChatGPT to rewrite an email.
The gap Sam Altman keeps trying to name is not a capability gap. It's a metabolism gap.
A model version ships in 8 weeks. Procurement in a Fortune 500 takes 9 months. Retraining a workforce takes 3 years. The median company is 6 months into a pilot, 2 layers of legal review deep, running a model from two releases ago.
By the time the average enterprise finishes its "AI transformation," the model is already three generations obsolete. The certification treadmill starts spinning before the first workflow even launches.
Altman's second line is the one nobody clipped. What happens when these systems start improving themselves. Nothing in the current economy is built to absorb quarterly self-improvement cycles.
The models are already improving themselves. The world is still learning to prompt.
Save this before the next AI budget your CFO signs.
Google invented the Transformer. Google built a ChatGPT-level bot two years before OpenAI. Google refused to ship it. A co-creator of the architecture just explained why
Noam Shazeer built Meena (later LaMDA) inside Google. Tens of thousands of Google employees were already using it internally and loved it. Leadership kept it behind locked doors.
The unspoken fears: "What if people stop using Google Search?" and "What if it says something harmful?"
Both concerns missed the point. A chatbot that answers questions is meant to reduce search queries. A model creative enough to be useful is also complex enough to hallucinate. That’s the trade-off.
Google created the architecture. Google trained the model. Google had a massive head start on the exact product OpenAI would later ship to 100 million users.
They sat on it because launching it threatened a search revenue machine that was destined to be disrupted anyway.
The Innovator's Dilemma in a single sentence: The incumbent has the technology, the talent, and the distribution. The only thing they lack is the willingness to eat their own lunch.
Save this for the next time you catch yourself protecting a legacy product from its obvious successor.
Follow if you'd rather build the future than protect the past.
@0x_freddy Orchestration > Prompting. The teams adapting to this early will operate with 1/10th the headcount and 10x the output
Anthropic just published the playbook for automating an entire company using AI agents. The engineers who built Claude wrote it. 37 minutes. Free. Nobody outside Anthropic has been able to write something this specific
If you run a team of five or more in 2026, half of what those people do is on the whiteboard in this video. Not "will be" replaced. Is being replaced. This week.
The video walks through what a working agent stack looks like on the inside. Agents that split tasks between themselves. Agents that check each other's output. Agents that escalate to a human only when everything else has failed. The pattern nobody selling a "prompt engineering" course knows how to demo.
Every consultant billing $50,000 to talk about "AI transformation" is about to become optional. Their entire deck is a longer version of what Anthropic just released for free.
You have two options. Watch the 37 minutes and rebuild your team's workflow next quarter. Or wait six months and pay a competitor who already did.
Save this before the next off-site strategy meeting.
Sam Altman admitted what everyone in Silicon Valley was thinking: Google should have completely dominated the AI race from day one
Google invented the Transformer architecture in 2017. Then, every single one of the eight authors on that seminal paper left.
Noam Shazeer left to start Character-AI. Aidan Gomez founded Cohere. Ashish Vaswani and Niki Parmar started Essential AI. Jakob Uszkoreit founded Inceptive. Llion Jones founded Sakana AI. Lukasz Kaiser joined OpenAI. Illia Polosukhin built NEAR Protocol.
Google had the paper, the compute, the users, and the distribution. What they lacked was a culture that allowed the people who invented the tech to actually ship it.
The talent didn't leave for higher salaries. They left because bureaucracy slowed down execution. Within eighteen months of leaving, every single author was shipping groundbreaking models somewhere else.
The lesson goes far beyond Google. It’s what happens to any incumbent when protecting legacy revenue becomes more important than shipping the future.
Sundar Pichai kept the CEO chair. The eight engineers who wrote the paper built the rest of the AI industry.
Follow if you'd rather build for the future than protect the past.
For 27 years, everything Google knew about building AI stayed inside Google. Jeff Dean just put all of it online
One hour. Free. Nobody else on the planet could give this lecture.
Dean is Google's Chief Scientist. He was employee number 25. He built MapReduce, BigTable, TensorFlow. Every piece of infrastructure that made modern AI possible has his fingerprints on it. When he compresses his career into one hour, you watch.
He walks the whole stack. LLMs. Prompts. Agents. Agent teams. Graphs. The final section is the one that matters: one human coordinating 100 agents at once. That's how the top team at Google actually ships now.
You're here for one of three reasons. Either you already run agent systems and want to see what the next tier looks like. Or you're still writing prompts and haven't noticed the tier above you exists. Or you're managing engineers who go redundant if they don't watch this too.
The hour is free right now. In six months every senior AI role will require what Dean covers between minute 52 and 60.
Watch it before your team's next planning meeting.
https://t.co/eflBjoVyV2
Top hedge fund analysts get sent one audio recording on their first day at a new desk. Not a book. Not a model. A masterclass by a 99-year-old billionaire listing every way your brain will lose you money
He sat down and covered 25 biases in one sitting. Some still aren't in any MBA curriculum. Business schools charge $200,000 to teach a subset of what he gave away in that room.
He didn't sell a course. He didn't write a newsletter. He walked in, sat down, listed the traps, and walked out.
The uncomfortable part was who was in the room. Money managers. He told them index funds will crush most of them. He called market manias antisocial. He said speculative hype is full of wretched excess. Nobody argued.
40 million people have heard his name. Almost none of them have listened to him walk through the 25 ways their own brain is working against them.
The lecture is free. He passed away at 99.
Send this to the next person who tells you they "did their own research."
Google invented the Transformer. Google built a ChatGPT-level bot two years before OpenAI. Google refused to ship it. A co-creator of the architecture just explained why
Noam Shazeer built Meena (later LaMDA) inside Google. Tens of thousands of Google employees were already using it internally and loved it. Leadership kept it behind locked doors.
The unspoken fears: "What if people stop using Google Search?" and "What if it says something harmful?"
Both concerns missed the point. A chatbot that answers questions is meant to reduce search queries. A model creative enough to be useful is also complex enough to hallucinate. That’s the trade-off.
Google created the architecture. Google trained the model. Google had a massive head start on the exact product OpenAI would later ship to 100 million users.
They sat on it because launching it threatened a search revenue machine that was destined to be disrupted anyway.
The Innovator's Dilemma in a single sentence: The incumbent has the technology, the talent, and the distribution. The only thing they lack is the willingness to eat their own lunch.
Save this for the next time you catch yourself protecting a legacy product from its obvious successor.
Follow if you'd rather build the future than protect the past.
Google invented the Transformer. Google built a ChatGPT-level bot two years before OpenAI. Google refused to ship it. A co-creator of the architecture just explained why
Noam Shazeer built Meena (later LaMDA) inside Google. Tens of thousands of Google employees were already using it internally and loved it. Leadership kept it behind locked doors.
The unspoken fears: "What if people stop using Google Search?" and "What if it says something harmful?"
Both concerns missed the point. A chatbot that answers questions is meant to reduce search queries. A model creative enough to be useful is also complex enough to hallucinate. That’s the trade-off.
Google created the architecture. Google trained the model. Google had a massive head start on the exact product OpenAI would later ship to 100 million users.
They sat on it because launching it threatened a search revenue machine that was destined to be disrupted anyway.
The Innovator's Dilemma in a single sentence: The incumbent has the technology, the talent, and the distribution. The only thing they lack is the willingness to eat their own lunch.
Save this for the next time you catch yourself protecting a legacy product from its obvious successor.
Follow if you'd rather build the future than protect the past.
Boris Cherny leads Claude Code at Anthropic. He just spent 31 minutes explaining how top engineers are actually building software today.
His core premise: You're not supposed to write code line-by-line anymore. You're supposed to design autonomous systems that handle the loop.
The real shift is moving from single prompts to multi-agent architectures. One agent writes, one reviews, one tests, and one manages the deployment. Each with clear boundaries, checkpoints, and handoffs.
Most developers know how to prompt an AI for a single function. Almost nobody knows how to chain agents into an automated workflow that runs without hand-holding. That’s the real skill gap in 2026.
He broke down the exact patterns used internally at Anthropic: how to manage agent memory, when to isolate context, and how to prevent agents from undoing each other's work.
Skip Netflix tonight. Watch how the creator of Claude Code actually builds software.
Follow if you'd rather architect AI systems than manually fix code the model could solve for you.
She built 100+ agents at Anthropic and cleared $1.3M in the process. She just spent 60 minutes at Stanford giving away exactly how
- 02:12 how her first agent at Anthropic cleared $1.3M
- 08:40 the agent that replaces a $200K/month engineering team
- 19:52 one overnight run that closed a 5-year roadmap
The three things nobody teaches in a tutorial: what to give an agent memory of, what to never let it decide alone, and how to know when it has learned something worth keeping.
Most engineers ship one agent, watch it work once, and call it done. The playbook she's built at Anthropic assumes the first version is the worst it will ever be. Every run feeds the next one. That's the difference between a demo and a system.
The Stanford room walked in thinking they'd learn how to prompt. They left with a spec for building software that gets smarter without them.
Save this before the next person you talk to says "I've been meaning to try agents."
https://t.co/LVxrtk9baX
In the 70s, game theory was hyped the way AI is now. Every strategist was obsessed with one question: how to win
Then a researcher watched his kids play legos and realized the field was solving the wrong problem.
Kids don't play legos to win. There's no scoreboard. Whoever gets bored first loses. That turns out to be how business actually works.
The mistake is treating an infinite game like a finite one. Finite games end with a winner. Infinite games have shifting players, evolving rules, and one goal: keep playing.
Every company that ever declared it "beat the competition" was dying by the time it said it out loud. Blockbuster beat Netflix in 2004. Kodak beat digital in 1996. The scoreboards were real. The scoreboards were wrong.
Simon Sinek's walkthrough is cleaner than any business book this year. He shows what an infinite mindset looks like in practice, and why the companies you admire will lose to the ones playing to stay in the game.
Save it for the next time you catch yourself trying to win the wrong game
https://t.co/YVS3eegciV
Boris Cherny leads Claude Code at Anthropic. He just spent 31 minutes explaining how top engineers are actually building software today.
His core premise: You're not supposed to write code line-by-line anymore. You're supposed to design autonomous systems that handle the loop.
The real shift is moving from single prompts to multi-agent architectures. One agent writes, one reviews, one tests, and one manages the deployment. Each with clear boundaries, checkpoints, and handoffs.
Most developers know how to prompt an AI for a single function. Almost nobody knows how to chain agents into an automated workflow that runs without hand-holding. That’s the real skill gap in 2026.
He broke down the exact patterns used internally at Anthropic: how to manage agent memory, when to isolate context, and how to prevent agents from undoing each other's work.
Skip Netflix tonight. Watch how the creator of Claude Code actually builds software.
Follow if you'd rather architect AI systems than manually fix code the model could solve for you.
https://t.co/RbVSSvaqXq
A 19-year-old girl bills around $9,000 a week making websites she doesn't know how to code. Six months ago she was making $15 an hour at Starbucks. Her dad still thinks she works there
She opens Google Maps, types "tacos philadelphia," picks a 4.7-star spot with a website from 2013, drops the reviews and menu into Lovable, hits generate. Twenty minutes later a modern site with the taqueria's name on it lands in a preview link.
The owner opens the link. Shows his daughter. "Is this us?" He pays $850 that evening.
Month one she cleared $600. Month three she was billing $4,000 a week. Last week: just over $9,000.
An agency would quote the same job at $12,000, four people, six weeks. She does it alone in 20 minutes with zero investment.
You are not competing with agencies anymore. You are competing with a teenager who cracked the search that closes a deal in 20 minutes.
The next $10,000-a-week freelancers won't come from bootcamps. They'll come from whoever notices which query was worth $850 all along.
como crear paginas web ganadoras de $50,000 con código de Claude + Fable 5
esto es un tutorial completo de 25 minutos
en un solo video, os lo enseñan paso a paso.
@uday_devops Impossible
@ToolySOL Yes, and memcoin trading on Solana
@piyush784066 Depends on the problem that needs to be solved. For example, I would choose ChatGPT
@uday_devops Codex or Claude😁
@elonmusk Fable 5 is #1 on paper, but Anthropic just crippled it with safety classifiers and loop limits. Grok 4.5 at #2 without the guardrails is the actual winner for autonomous production
@Mr_Salio They restarted training because DeepSeek and Qwen are eating their lunch for pennies
By the time this ships, a kid in Shenzhen will be running a better model locally on a Mac Mini.
@Monika2r Amateurs pick favorites. Pros build routers
Claude for the brain, DeepSeek for the bulk, Perplexity for the facts. Single-model setups are why your API bill is bleeding
@OpenAIDevs The era of using one model for everything is over.
If you default to Sol, you're just burning margin. Routing is the new prompt engineering.
@OpenAI It's cool that AI technologies are finding such applications
Last Seen Users on Sotwe
Most Popular Users
Elon Musk
@elonmusk
241.3M followers
Barack Obama
@barackobama
119.1M followers
Cristiano Ronaldo
@cristiano
113M followers
Donald J. Trump
@realdonaldtrump
111.8M followers
Narendra Modi
@narendramodi
107.1M followers
Rihanna
@rihanna
98.3M followers
NASA
@nasa
92.3M followers
Justin Bieber
@justinbieber
91.5M followers
KATY PERRY
@katyperry
89.1M followers
Taylor Swift
@taylorswift13
83M followers
Lady Gaga
@ladygaga
74.5M followers
Virat Kohli
@imvkohli
72.1M followers
Kim Kardashian
@kimkardashian
70.5M followers
YouTube
@youtube
68.8M followers
Neymar Jr
@neymarjr
65M followers
Bill Gates
@billgates
64.7M followers
The Ellen Show
@theellenshow
62.4M followers
Selena Gomez
@selenagomez
62.2M followers
CNN
@cnn
61.8M followers
X
@x
60.8M followers