Richman

Most people read the news - I trade it

[Richman @RichmanNYC]

🚨 WILD: A Polymarket trader turned $13 into $4,068 — just by betting on the weather.

He bought YES at 0.2¢ on “Madrid high ≤17°C on May 7” and watched it hit.

This isn’t a one-off either. RamsBR has been quietly farming weather markets and is up $15k+ in the past month alone.

[Richman @RichmanNYC]

@HarlowSpectre Gm Harlow, insane interior btw

[Richman @RichmanNYC]

@HarlowSpectre @ciindae Nike Pro

[Richman @RichmanNYC]

A 24 year old in Stockholm runs modern AI on a 1999 iMac. No internet. No subscription. No cloud account. The model responds instantly on hardware older than most of the people reading this post.

The setup is a translucent blueberry iMac G3 her parents kept in storage for 27 years. PowerPC processor. 128 MB of RAM. The kind of machine that loaded a single webpage with a coffee break in 1999. She plugged it into a local network, pointed it at a small open source model running on a separate box in the same room, and started chatting through Safari.

The model answers in under a second. Every prompt processed in the apartment she’s sitting in. Nothing leaves the building. Her electricity bill went up by maybe two dollars.

The premise nobody questions is that AI requires a data center. It doesn’t. Data centers exist because OpenAI, Anthropic, and Google serve billions of simultaneous requests. You are not billions of users. You are one person with maybe fifty prompts a day. A processor older than your little brother handles that without breaking a sweat.

Her version isn’t as smart as GPT-5. It doesn’t need to be. It answers her work emails, summarizes her notes, drafts her cover letters, and writes her code without ever sending a single keystroke to a server she doesn’t own.

The wild part is the math. A used Mac Mini costs $200. A small open source model costs zero dollars. A power bill increase costs less than a cup of coffee. The total monthly cost of her stack is rounding error compared to a single ChatGPT Plus subscription.

The cloud AI industry is selling backup generators to people who already have working outlets in the wall. Most users will keep paying because the outlets are hidden behind a door labeled “complicated” that nobody bothered to open.

She opened it. The lights came on. The bill stopped.

[Richman @RichmanNYC]

A 16 year old in Bangalore made $10,000 from one YouTube channel last month. Now he’s running dozens of them from the same laptop. His mom thinks he’s doing homework.

The setup is brutally elegant. An antidetect browser isolates each identity so YouTube can’t link the accounts. Automation handles uploads on a schedule. The same content pipeline gets copy-pasted into a new niche every week. Animals on Monday. Cars on Tuesday. Horror, sleep stories, true crime, motivation, anime tier lists, recipe shorts. Every channel gets its own backstory and posting time and looks like a totally fresh creator.

Each new channel costs $57 a month to operate. That covers the proxy, the antidetect license, the cloud API for the AI generation, and the upload automation. Past 30 days into the algorithm and YouTube starts pushing the content to organic feeds like any other channel.

The math is offensive once you see it. He runs forty channels. Half of them die in obscurity. Ten of them break even. Five of them hit a few hundred thousand views a month. One of them blows up. The one that blows up pays for the other 39 a hundred times over.

This stopped being a technical problem six months ago. It became a niche selection problem. The technology is solved. The infrastructure is renting for less than a Netflix subscription. The only edge left is picking the topic that the algorithm hasn’t seen six thousand times yet.

The wild part isn’t the money. It’s the asymmetry. A kid who isn���t legally allowed to open a bank account is outearning every middle manager at the local Infosys office, working from a laptop his uncle bought him for his last birthday.

The old model was one person makes one channel. The new model is one person runs a small media company while pretending to do their geometry homework.

The platforms haven’t caught up yet. The kids already did.

[winkle. @w1nklerr]

https://t.co/qVlSyWCYsM

[Richman @RichmanNYC]

@HarlowSpectre Follows done, cant wait for the guide drop 🤙

[Richman @RichmanNYC]

A 19 year old in Lagos makes $9,000 a month from cartoons she never drew, in a voice she never recorded, on a phone that fits in her back pocket.

The workflow is almost embarrassingly simple. Find a viral kids song on YouTube. Paste the lyrics into ChatGPT and ask for a version with the same rhythm but different words. Drop one line into Sora 2 Pro about colorful characters dancing for a kids show. Wait six minutes. Upload to YouTube Shorts. Repeat tomorrow.

Total time per video: under 20 minutes. Total cost: a $20 subscription and a phone she already owns. No camera. No mic. No laptop. No face on screen ever.

The math is what makes the room go quiet. Kids content has some of the highest CPMs on YouTube because parents loop the same video forty times before bedtime and advertisers pay a premium for the captive attention. One faceless channel pulling a few million views a month is $9K in ad revenue. Two channels is rent in any major city in the world.

Here’s the part nobody wants to say out loud. There’s nothing in this workflow that requires her. No talent. No experience. No special knowledge. The same steps work for you, for your cousin, for the kid sitting next to you on the train right now.

The window is brutally short. The moment YouTube tightens its policy on AI generated kids content, or every creator on Earth piles into the niche, the CPM collapses and the channels stop printing. Six months. Maybe twelve.

Most people will read this, nod, and do nothing. The ones who open Sora tonight will be eating dinner poolside by Christmas.

The gap between knowing and doing has never been this expensive.

[Richman @RichmanNYC]

@HarlowSpectre Absolute Cinema

[Richman @RichmanNYC]

Apple spent $30,499,990 and four years of engineering to put your face on the outside of a headset. A guy in Tokyo did the same thing this weekend for $8.

The product is a lapel pin. A tiny LED panel clipped to his jacket that mirrors his face in real time. He smiles, the panel smiles. He blinks, it blinks. The latency is so low it reads like a mirror instead of a recording. Total bill of materials: $8. Runs on a small lipo battery you can charge with a USB-C cable.

Apple’s version of this is called Persona. It needs an M2 chip, an R1 co-processor, 12 cameras, a LiDAR sensor, a TrueDepth array, a lenticular OLED that splits the image into two eyes to fake 3D depth, and a proprietary 13 volt cable they invented so you can’t accidentally plug it into anything else. Hundreds of engineers. Years of R&D. $3,499 at checkout.

The Japanese guy ran MediaPipe on an ESP32. Three dollar microcontroller, off the shelf, ships from Shenzhen overnight.

This is the part of every tech cycle nobody wants to admit out loud. The flagship feature of a $3,500 spatial computer just got cloned in a hoodie pocket by one person over a weekend. Apple calls it Persona. He calls it a badge.

The gap between “needs a dedicated spatial operating system” and “runs on a $3 chip” is supposed to take decades to close. It took 72 hours.

Every Vision Pro engineer woke up today knowing the moat is now a puddle. The next ten years aren’t about who can build the most complex stack. They’re about who notices the cheapest version got good enough.

[Richman @RichmanNYC]

@HarlowSpectre Gm Harlow, have a nice day ☕️

[Richman @RichmanNYC]

A trader bought a memecoin at $2M market cap, knew it was a hack, and still walked away from a $30K bag because the chart “didn’t feel great.”

His name is Ash Robin. The token was the Roaring Kitty memecoin that briefly hit $10M after his account got compromised. Ash was online in the exact minute the post went up. Bought $7K at an average market cap of $2M. Held while it dipped to $1.5M. Sold flat.

Then two community teams took it over. CTO’d it. Pumped it from $2M to $8M. He watched from the sidelines with a sold bag and his cursor on the buy button.

His own quote: “I bottom ticked it. Had a great entry. Did not hold long enough.”

The wild part isn’t the loss. It’s that the entry was perfect. The thesis was right. The timing was right. The exit was the only thing he got wrong, and the exit is the entire trade.

Every trader’s career has the same recurring villain. It’s not the bad call. It’s the right call sold too early. Conviction is the only edge that compounds. Everyone has the thesis. Almost nobody has the stomach.

He ended with “no risk no Ferrari am I right.” The Ferrari was already at the door. He just got out of the car too soon.

[Richman @RichmanNYC]

@PathOf_Legacy Benz 100%

[Richman @RichmanNYC]

@lynk0x Better than nothing btw

[Richman @RichmanNYC]

@quantgz Congratulations quant 🤘

[Richman @RichmanNYC]

A guy spent $55 on a Raspberry Pi and shipped a $740 per month SaaS by Monday morning.

The product is brutal in its simplicity. Small Shopify stores hate checking competitor prices by hand every week. The owner used to spend Sundays in a spreadsheet copying prices off three websites.

Now a Pi the size of a deck of cards does it in the background and emails them a Monday morning report.

Total stack: $55 hardware. Claude Code wrote the scraper and the dashboard in a weekend. Stripe handles billing. Supabase holds the data. The whole thing lives in his living room.

Pricing is $49 a month per store. 15 stores signed up in the first three weeks. $735 MRR, growing by word of mouth between Shopify owners in the same Slack group.

The boring part is the moat. Anyone could build it. Almost nobody will. Because the idea sounds too unsexy to chase. No demo day judge gets excited about a Sunday spreadsheet for a candle store in Ohio.

Every gold rush has two kinds of people. The ones hunting the giant vein. And the ones quietly selling shovels next to the trail.

Most people are still asking Claude to write them a poem. He pointed it at one annoying weekly task and built a factory.

The opportunity isn't AI. The opportunity is the boring task next to AI.

[Richman @RichmanNYC]

Dana White woke up thinking he was down $90K. The actual number was $3 million.
He didn’t even know.

Drove off the next morning, called the casino host to brag about coming back to win his $85K back. The guy on the other end of the line goes quiet for a beat. “Dana, you lost three million last night.”

Dana pulled over on the side of the road. His personal limit is one million per night. The casino had to call the General Manager at 4am to extend it. Dana had been on the phone with the GM personally, calling him names until he caved.

He doesn’t remember any of it. The casino reminded him. He said yeah, sounds like me.

The wild part isn’t the loss. It’s the gap. $90K and $3M felt the same to him in the morning. Most people will work three lifetimes to earn what he forgot losing in one night.

Different gravity on different planets. He lives on one where three million dollars doesn’t dent the calendar.

[Richman @RichmanNYC]

@RaminNasibov That’s how i look at my 1 like on a post

[Richman @RichmanNYC]

@FearedBuck Still in Prime btw

[Richman @RichmanNYC]

@HarlowSpectre Gm Harlow, have a nice weekend 🤙

[Richman @RichmanNYC]

@SaulWgmi Never worked again

[Richman @RichmanNYC]

@WatcherGuru X is for Men