Marrow (@marrownets) header
Marrow (@marrownets) profile picture
Joined September 2026
Photo by Marrow (@marrownets)Photo by Marrow (@marrownets)Photo by Marrow (@marrownets)Photo by Marrow (@marrownets)Photo by Marrow (@marrownets)Photo by Marrow (@marrownets)Photo by Marrow (@marrownets)Photo by Marrow (@marrownets)Photo by Marrow (@marrownets)Photo by Marrow (@marrownets)
Marrow retweeted
This is f*cking insane. Elon Musk just described how Grok decides what you see on X, and the scale is unreal: "You take the top 10 million posts a day, understand every one, and match them to the person who'd actually want them" In this October 2025 conversation he lays out the whole plan, from reading the feed to searching it by meaning The starting point is the top 10 million of roughly 100 million posts a day. Read them, understand them, categorize them, match them to people. Then scale up to the whole stream. The question becomes: out of everything posted today, what would actually interest you? Asked how much compute that takes, he puts it around 50,000 H100s. Then it moves to search. You ask in plain language, and it finds the text, images or video that match the meaning, not the keywords. He's describing the plan, hardware and all. Whether you want a model reading everything you post is a separate question. Read the Feed → Understand It → Match It to You → Search by Meaning The feed was never a list of who you follow. It's a model deciding what you'd want to see. 95% will argue about the algorithm, the 5% who build on this kind of system will own what it surfaces Save this and send it to anyone who thinks the feed is still chronological then read the article below ↓
8
6
50
3,665
Marrow retweeted
This is f*cking insane. Elon Musk was asked the secret to getting rich. His answer was two words: "Don't chase money. First, create more value than you receive" In 2 minutes he takes apart why chasing the sale is backwards Hard work alone isn't enough. Sales are not the goal, they're a result, one measure of how much value you delivered So you don't start with how much you want to make. You start with whose problem you can solve, and whether they'd pay to have it solved Before you chase a single sale, answer four things: - whose problem are you solving - what outcome can you deliver - why should they buy it from you - how much is that worth to them Their Problem → A Better Solution → Value Delivered → A Reason to Pay → Money as the Result You don't get paid for effort. You get paid for solving someone's problem 95% chase the sale, the 5% who build something people genuinely need are the ones who keep making money for years Save this if you're starting anything this year then read the article below ↓
3
1
36
2,273
Marrow retweeted
This is f*cking insane. Elon Musk wants to launch AI satellites off the Moon, and the verb he uses gives the whole plan away: "You don't launch them off the Moon, you shoot them into space" In under 2 minutes he explains why the Moon changes what can do the launching On Earth a launch fights its way through the air and climbs out of our gravity well. Every kilo costs a rocket. The Moon has no atmosphere and about a sixth of the gravity. That changes the math completely. His pitch is a linear electric motor laid on the lunar surface. It accelerates the satellite and lets go. The satellite leaves, the expensive launcher stays put for the next one. The rocket guy's Moon plan skips the rocket entirely. He's only giving the upside. The build, the power and the cost of getting the motor to the Moon in the first place are a separate problem. Earth Rockets → No Atmosphere → Low Gravity → A Launcher That Stays → Satellites Shot From the Moon The rocket was never the point. The reusable launcher that never leaves the ground is. 95% will argue about whether it's possible, the 5% who build for the next decade already see where this goes Save this and send it to anyone who thinks space still means rockets then read the article below ↓
4
26
1
187
30,296
Marrow retweeted
This is f*cking insane. Elon Musk just put a number on what it costs to build a humanoid robot: "At a million units a year, Optimus could cost about $20,000 to make. The swing factor is how cheap the AI chip inside it gets" In 2 minutes, he turns the robot story from a price tag into an economics problem The $20,000 is not a sticker price. It's the marginal cost to manufacture, and only once output reaches a million robots a year Then he names the one variable that moves everything: how much the AI chip inside the robot costs So the real question is not how cheap the robot is. It's how cheap the compute inside it gets He's describing the economics he hopes to reach, not a robot you can order at that price today Chips → Scale → $20k Robot → Physical Labor → The People Who Build On Top The robot is just the body. The real money is in the compute inside it and what you build on top 95% will argue about the price tag, the 5% who build on the compute will own what it produces Save this and send it to anyone who still thinks robots are a hardware story Then read the full Grok Bot Playbook below ↓
15
15
2
144
112,024
Marrow retweeted
This is f*cking insane. Elon Musk needed 100,000 GPUs training together. Suppliers told him 18-24 months. He said they needed it done in six. Otherwise, xAI would not be competitive. In this June 2025 interview, he explains what stood in the way. They found an old Electrolux factory in Memphis. It had 15 megawatts of power. They needed 150. So they rented generators. Then, he says, roughly a quarter of America's mobile cooling capacity. Still not enough. During training, power demand could drop by 50 percent in 100 milliseconds. The generators could not respond fast enough. So they added Tesla Megapacks and tweaked the software to smooth the load. Then came the networking: making 100,000 H100s train together. Four shifts. Around the clock. Musk says he slept in the data center and did cabling himself. The "AI race" sounds abstract until you hear about the generators, cooling gear, batteries and cables behind it. Save this breakdown. Then read the full Grok Bot Playbook below ↓
2
24
1,724
Marrow retweeted
This is f*cking insane. Elon Musk was asked about his worst habit. The man building rockets gave the most ordinary answer possible: "I end up checking my phone first. But I think it's a bad habit" In under a minute, the richest guy in the room admits he loses the same morning fight you do He knows it's bad. He wants to swap it for 20 minutes of exercise. He also admits he doesn't really like exercising His real point lands harder than the joke: running a company and fixing one small daily habit are two completely different skills You can look like a total success to everyone else and still have the same flaws you can't shake Sometimes changing those 20 minutes in the morning is harder than changing the world Phone → Habit → A System → Agents → The Routine That Runs Itself Willpower loses this fight every time. A system that runs the routine for you doesn't 95% will keep blaming their willpower, the 5% who build a system that runs it won't need any Save it and send it to anyone who grabs their phone before their feet hit
2
1
1
25
2,227
Marrow retweeted
This is f*cking insane. Elon Musk was asked why he keeps building AI data centers that half the town shows up to protest. He answered with one city: "We're paying about twice the area's previous wage levels" In 2 minutes, he takes the whole data-center fight apart with one example: Memphis There was real opposition. Then the facility opened Now the restaurants are packed. The hotels are fully booked. Local businesses are thriving Wages roughly doubled. The company funds local schools. The city and county pull in serious tax revenue His point is simple: before it's built, people picture only the downside. After it's built, the money starts moving to businesses that have nothing to do with the data center He's only giving the company's side. The power bills, the air and the noise still land on residents. Those are separate questions Energy → Data Center → Jobs → Local Boom → The People Who Build On Top The data center is just the shovel. The real money is in what you build on the compute 95% will keep debating the power bill, the 5% who build on the compute will own what it produces Save it and send it to anyone watching the AI buildout from the sidelines Then read the full Grok Bot Playbook below ↓
53
198
10
927
219,660
Marrow retweeted
This is f*cking insane. Elon Musk said SpaceX paid $140,000 over water used to cool a rocket launch pad. His account starts at Starbase, in a tropical thunderstorm region. He describes the water as drinking water. Says it caused no harm. Points out that water falls from the sky there all the time. Then comes the permit. "because you don't have a permit" Musk says the next step was worse: further Starship launch applications would not be processed until the fine was paid. So SpaceX paid it. Cooling water → missing permit → $140,000 → the next launch application. His closing point is Mars: at that pace, he says, they will never make it there. 83 seconds. Watch him walk through the dispute and decide what you make of the trade-off. Watch it, then read the full Grok Bot Playbook below ↓
2
3
36
24,622
Marrow retweeted
Holy f*ck. SpaceXAI senior engineer Lauren Tan: "Grok Bot is the most powerful agent tool we've ever released, and only 1% of users know how to use it." In this 40-minute session, she shows how she set up her team before she scaled it Job → Chief of Staff → 3 Managers → 11 Workers → 200+ Agents* Most people add another bot once the first one fails. That simply multiplies the mistakes Give every bot one job and a manager that checks it, instead of paying for the same failure each run Worth more than most $1,000 agent engineering courses Bookmark and watch it today Then read the full Grok Bot Playbook below ↓
3
1
18
1,036
Marrow retweeted
This is f*king insane. SpaceXAI lead engineer (ex-Cursor): "99% of people still use Grok Bot like a search bar. I run 20+ cloud agents with a Chief of Staff, a PM and 15+ workers, and none of them ever clocks out" In this 22-minute session, he shows how he turned one bot into a team that works while he sleeps Grok Bot → Chief of Staff → PM → 15+ Workers → 200+ Agents* Most people open a second chat once the first one stalls. That's only two bots waiting for you Give the team a Chief of Staff that plans the day and a PM that checks each handoff, and you stop paying for work nobody reviewed Worth more than most $1,000 agent engineering courses Bookmark and watch it today Then read the full Grok Bot Playbook below ↓
3
1
16
526
Marrow retweeted
This is f*cking insane. SpaceXAI senior engineer Lauren Tan: "Grok Bot is the most powerful agent tool we've ever released, and only 1% of users know how to use it." In this 40-minute session, she shows how she set up her team before she scaled it Job → Chief of Staff → 3 Managers → 11 Workers → 200+ Agents* Most people add another bot when the first one fails. That just multiplies the mistakes Give every bot one job and a manager that checks it, instead of paying for the same failure every run Worth more than most $1,000 agent engineering courses Bookmark and watch it today Then read the full Grok Bot Playbook below ↓
2
1
14
354
Marrow retweeted
This is f*cking insane. Elon Musk just told his entire SpaceXAI team: "Maybe even by the end of this year, you don't even bother doing coding. The AI just creates the binary directly It should be possible to completely emulate any company where the output is digital" In this 42-minute all-hands, he breaks down what happens to engineers and to every business that runs on a screen Code → AI Code → No Code → Companies Run By AI His coding team claims Grok Code is already training its own next version If your work happens on a screen, watch this before the year ends Worth more than any $3,500 AI course Save it and watch it tonight Then read the full Grok Bot Playbook below ↓
1
16
647
Marrow retweeted
This is f*cking insane. SpaceXAI engineer Nick Miller (ex-Cursor): "99% of people open one Grok Bot and call it automation I run 30+ bots with a Chief of Staff, a PM, QA and 20+ workers" In this 30-minute demo, he grows one bot into a team that passes work along and checks itself Grok Bot → Chief of Staff → PM → QA → 30+ Bots, One Budget Most people assume more bots means a bigger bill. The bill comes from bots re-reading everything Give every bot one job and a lean context, and the team costs less than one bloated chat Worth more than most $1,000 agent engineering courses Bookmark and watch it today Then read the full Grok Bot Playbook below ↓
1
19
282
Marrow retweeted
This is f*cking insane. SpaceXAI engineer: “SpaceXAI pays me $1.2M a year because i know how to build 150+ GrokBot teams that ship 1200+ updates a day while i sleep. I haven’t written code by hand in months. I set the rules, then check what the bots deliver.” In a 1-hour livestream, she begins with an empty GitHub and zero bots. 72 hours later, her team has built a company run by bots. Here’s the system: step 1 → give each bot one clear role and its own computer in the cloud. step 2 → teach it a task once. it stores the process as a reusable skill. step 3 → put the bots in one group chat so they can pass work to each other. step 4 → set the rules, let the team run, then review what gets done. Even the 4 a.m. run goes on with your laptop closed. GrokBot → skills → bot team → one group chat → a company that runs itself SpaceXAI calls it “a persona, not a task.” Anybody can open a new GrokBot chat. the real unlock is building a team that keeps going without you. This 1-hour tutorial is worth more than most $2,000 agentic engineering courses. Bookmark it and watch today Then read the full Grok Bot Playbook below ↓
3
27
704
Marrow retweeted
This is f*cking insane. SpaceXAI senior engineer (ex-Cursor): "I run 60+ Grok Bots like a company. A Chief of Staff runs my day, 3 PMs assign the work, 50+ workers ship it overnight" In this 1-hour workshop, he sets up that team from scratch, step by step Job → Chief of Staff → PMs → 50+ Workers → 60 Bots, One Budget Most people think 60 bots means 60x the bill. It doesn't, if every bot only reads what its role needs Pay for clear roles once, instead of paying every bot to re-read the whole chat Worth more than most $2,000 agent engineering courses Bookmark and watch it today Then read the full Grok Bot Playbook below ↓
1
1
15
395
Marrow retweeted
This is f*cking insane. SpaceXAI founder Elon Musk: "Grok 4.6 is the #1 coding model right now, and in 6-12 months GrokBot will be the #1 agentic tool in the world At SpaceXAI, over 70% of our engineers already use GrokBot to build agents that learn by themselves" In this 1-hour session, Elon sits down with SpaceXAI engineers and lays out what AI engineering looks like next Prompts → GrokBot → Agents → Self-Learning Systems → AI That Builds AI Using AI doesn't mean asking Grok a question and copying the answer It means agents that learn from every run and improve without you touching them 95% of people still use Grok like a search bar, the 5% who build self-learning agents will run everything Worth more than a Stanford course on AI engineering Bookmark and watch it today Then read the full Grok Bot Playbook below ↓
2
13
609
Marrow retweeted
This is f*cking insane. SpaceXAI engineer (ex. Cursor): "I can throw this laptop into the river, and my Grok Bots will keep working 24/7 one Bot as the CEO, the other 19 finish tasks and come back only with finished results" In a 15-min workshop, the SpaceX team shows how they build Grok Bots and how to really use these agents properly This replaces weeks of random agent tutorials Watch it, then read the full Grok Bot Playbook below ↓
1
21
296
Marrow retweeted
This is f*cking insane. SpaceXAI lead engineer (ex-Cursor): "99% of people still use GrokBot like Google. I run 20+ Grok cloud agents with a Chief of Staff, a PM and 15+ workers, and they never clock out" In just 20 minutes, this engineer shows how to turn one GrokBot into a team that works for you 24/7 GrokBot → Chief of Staff → PM → Workers → Autonomous Fleet A Chief of Staff agent plans the day and assigns every task A PM agent splits the work and reviews everything the workers ship Anyone can chat with one GrokBot, almost nobody builds the team that keeps working while they sleep Worth more than most $1,000 agentic engineering courses Watch it today, then read the article below ↓
1
22
550
Marrow retweeted
This is f*cking insane. SpaceXAI engineer Lauren Tan: "I was the human copy-paste between my agent and my browser. So I fired myself Now 30+ agents run around the clock, and my whole job is checking their work over morning coffee" In 1 hour, she shows step by step how she built that dark factory on Grok Bot Skills → Harness → Verification → Loops → Dark Factory Most engineers still babysit one agent. She quit that job Worth more than most $2,000 agentic engineering courses Bookmark and watch it today Then read the article below ↓
3
12
2
209
21,070
Marrow retweeted
This is f*king insane. SpaceXAI engineer Lauren Tan: "250 applications over two years. Not one offer Then he showed up running 20 agents and got a $500K offer" In this 1-hour workshop, Lauren shows how to build a Grok Bot team that delegates work and verifies the results Grok Bot → Chief of Staff → Managers → Workers A chat window gives you one bot. An agent team gives you a whole system This free workshop is worth more than most $800 agent engineering courses Bookmark and watch it today Then read the article below ↓
1
15
279