Smoooothly

Science|AI agents enjoyer

[Smoooothly @Smoooothly_]

@froster_one Is there any correlation here between the pandemic and the developing crypto market?

[Smoooothly @Smoooothly_]

@0xpupupu It's nice doc for your setup,a lot of space for lowcost

[pupupu @0xpupupu]

A hybrid array that runs 2 NVMe drives and spinning disks in the same box.

The ACASIS H004 takes two Samsung 990 PROs on top and 3.5-inch drives underneath. Fast storage and cheap storage in one enclosure, one cable, one power input.

The SSD side reads 1,463 MB/s over a single 40Gbps connection. The disks below handle everything that doesn't need to move fast.

That split is the whole point:

▸ Today's footage sits on the NVMe and plays back instantly.
▸ Everything older drops to the hard drives and still stays online.
▸ Nothing gets archived to a drive in a drawer.

Most people run an SSD on the desk and a stack of old drives in a box somewhere. This puts both in one enclosure and one cable.

[Smoooothly @Smoooothly_]

@0xFractaI bro really ran 100k simulations just to prove my intuition is trash bruh

[Smoooothly @Smoooothly_]

@froster_one u wanna say,i need to take my money and buy this stock?

[Smoooothly @Smoooothly_]

@0xpupupu i wanna get this to my setub ,it's so cutie

[pupupu @0xpupupu]

This is f**king gold.

A $12 board and an e-ink panel built the nicest thing on this desk.

It shows whatever Spotify is playing right now. Album art, artist, album, progress bar — pulled over wifi and rendered in black and white like a page out of a paperback.

Inside is an ESP32, a lithium pouch cell and a 3D printed shell. Nothing else.

What makes it work:

▸ No glow, no notifications, nothing competing with the monitor.

▸ The battery lasts weeks because e-ink holds an image on zero power.

▸ The screen only redraws when the track does.

Every desk has a plant, a lamp, a thing that does nothing useful. This is the version that costs $12 and looks better than both.

[pupupu @0xpupupu]

A 1984 Macintosh with a Mac Mini M4 living inside it.

The Wokyis shell is a scale replica of the original — beige plastic, rainbow logo, tiny CRT-shaped screen. The Mini slides in from underneath and the screen becomes a real second display.

An M.2 slot sits in the back. A Samsung 990 EVO Plus goes in at 4TB, mounting alongside whatever storage the Mini shipped with.

Then it stops being a toy:

▸ Card readers, USB and HDMI where the floppy drive used to be.

▸ The little screen runs macOS, not a loop of fake boot animations.

▸ One power input feeds everything.

Most docks hide under the machine and nobody looks at them twice. This one puts 4TB and a second screen on the desk and everyone in the room asks about it.

A dock, a display and 4TB of storage in the shape of the computer that started all of it.

[Smoooothly @Smoooothly_]

@0xpupupu if you have money,you need this
anyone will be delighted

[Smoooothly @Smoooothly_]

@HajiNoPrime @0xpupupu yeah,i want exactly that setup mimimi

[pupupu @0xpupupu]

A soldering iron just did what Apple charges $600 for.

The iMac shipped with 256GB. It's leaving the bench with 2TB, and Apple never sold that configuration at any price.

The storage chips are soldered to the board. Under a microscope they come off with hot air and bigger ones go down in the same pads.

• The chips cost about $40.
• Apple's price for the same jump is $600.
• 15x markup on parts anyone can order.

Nothing here was broken. The machine runs exactly like it did the first time it was switched on — it just came with the smallest configuration on the price list, and everyone accepted that as permanent.

The whole kit on the bench is hand tools. An iron, flux, hot air and a scope, and it costs less than one trip to the Genius Bar.

The iMac boots to the setup screen like it came out of a box. Same board, same year, 8x the storage.

$40 and an afternoon did what Apple sells for $600.

[Smoooothly @Smoooothly_]

@0xpupupu it's like a your personal ai agent literally wow

[pupupu @0xpupupu]

Bro fits a 67B AI model in a suitcase.

4 Mac Studios in a road case, wired over Thunderbolt 5, splitting one model across all 4 machines with EXO.

• 4 machines, one shared pool of unified memory.

• Thunderbolt 5 moves the layers between them fast enough that it works.

• Add a fifth Mac and the ceiling goes up again.

This is what you build when the work can't go through someone else's server. Medical records, legal discovery, client footage, anything under NDA — the model comes to the data instead of the data going to a datacenter.

The model runs on machines that draw less than a hair dryer between them, and the whole thing fits in the back of a car.

Every API call you'd make to run this costs money per token, forever. This one was paid for once and answers to nobody.

[Smoooothly @Smoooothly_]

A SCROLL-DRIVEN 3D SITE LIKE THIS QUOTES AT $8,000 WITH A 6-WEEK TIMELINE AND 3 PEOPLE ON IT.

A designer, a WebGL dev, a frontend engineer. 6 weeks because the 3D and the scroll logic get built twice, once in the mockup and once for real.

The stack underneath is not exotic. Three.js or React Three Fiber for the scene, GSAP ScrollTrigger to bind camera position to scroll offset, Lenis for the smooth scroll, custom shader materials for the glow and the volumetrics. Every piece is public and documented.

What made it expensive was the seam. The designer hands over static frames, the WebGL dev rebuilds them in code, and the 2 versions never quite match. Weeks disappear in that gap.

That seam is gone. Shader, scroll timeline and live browser output come out of the same pass, so there's no version to reconcile and nothing to hand across.

So one person now delivers in a week what took 3 people 6. The $8,000 doesn't get split anymore.

What still decides whether the site is worth $8,000 is the feel. Camera travel per 100px of scroll. Where a section snaps and where it drifts. Whether the glow reads expensive or reads like a bought template.

[Romario @bateshkaaa]

https://t.co/FCMzr9satN

[Smoooothly @Smoooothly_]

Freight agencies quote $18,000 and 6 weeks for a site like this. It was built in 1 night.

Claude Design does the art direction. Claude Code ships the build. Connect the two and the handoff between designer and developer stops existing.

No WebGL contractor. No motion studio. No 3 rounds of Figma before anyone writes a line of code.

What comes out the other side is a site a logistics group will pay $12,000 for without a second call. It looks like a company that moves 3,500 containers a month, not one that bought a template for $59.

That is the whole business now. Pick an industry that signs $40,000 contracts, build 1 site at this level, show it. The next 4 clients come from the first screenshot.

Design in 1 tool. Build in the other. Ship the same night.

Agencies still sell 6 weeks. You sell the finished thing.

[Romario @bateshkaaa]

https://t.co/FCMzr9satN

[Smoooothly @Smoooothly_]

Skip the market podcasts. One hour with Howard Marks and you'll stop asking the question that costs you the most money.

Marks co-founded Oaktree in 1995. He says the stupidest question in investing is whether we've hit the bottom.

His reasoning is hard to argue with. The bottom is the day before prices start rising again, so you can only recognize it the day after it's gone. Waiting for the bottom means waiting for something you'll never see in time.

He learned that in the only week that ever tested it.

September 2008. Lehman filed for bankruptcy. Bear Stearns, Wachovia, Washington Mutual and AIG were already gone. Oaktree sat on 10 billion dollars of raised, unspent capital while most people believed the financial system was about to stop existing.

There was no data to check. Nothing like it had happened before.

So Marks and his partner Bruce Karsh settled it in one sentence. If the world truly melts down, nothing they buy today will matter. If it doesn't and they sat still, they failed at their job.

That left one move.

They bought the most senior debt of companies that buyout firms had paid 4 billion for, at prices where the business could end up worth 1 billion and Oaktree would still get paid in full.

Prices kept falling. They kept buying. Karsh put in 450 million a week for 15 straight weeks.

Two things made it possible, and neither was a forecast. They raised the money before the crisis, because nobody funds you during one. And their own book was clean, so they spent the crisis buying instead of rescuing.

You can never tell when something has stopped falling. You can always tell when it's cheap.

Pay too much for a great company and you still lose.

[Kaien @kaienphase]

https://t.co/PbfztEwWL7

[Smoooothly @Smoooothly_]

@0xpupupu i want to try this setup but i broke(

[Smoooothly @Smoooothly_]

Skip the writing course. 4 minutes of Kurt Vonnegut with a piece of chalk will teach you more about story than any of them.

Everything you post competes on one thing. Does the reader stay. Vonnegut drew the answer on a blackboard.

He submitted it as a master's thesis at the University of Chicago. They rejected it, he said, because it was too simple and looked like too much fun.

The claim was that stories have shapes you can draw on graph paper.

Vertical line. Top is good fortune, bottom is sickness and poverty. Horizontal line across it, left is the beginning, right is the end. Plot how your character is doing along the way, and you get a curve.

The first curve he called Man in Hole. A person is fine, falls into trouble, climbs back out. Vonnegut said audiences never grow tired of it. You see it every day without noticing. The founder who went broke and came back. The guy who lost his job and now runs everything through AI. Same line, new costume.

The second is Cinderella. Bottom of the chart, then the godmother lifts her to the ball, then midnight collapses it. Not all the way down, though, because she still has the memory of the dance.

That midnight drop is the whole trick. The line falls, yet stays above where it began, so the reader keeps reading.

Then came the sentence the department did not want.

These shapes are simple enough to feed into a computer.

He said it in the age of punch cards. In 2016, researchers at the University of Vermont ran 1,327 books through a machine and mapped the emotional arc of each one. They found 6 shapes.

The rejected thesis was a description of the training data.

Before you write your next post, draw the line. If it runs flat, nobody is staying to the end.

[Smoooothly @Smoooothly_]

Instead of scrolling for 1 more hour, spend that hour learning what actually moves the price of everything you own.

John Geanakoplos teaches economics at Yale and runs money at a hedge fund at the same time. In 1997 he started writing about something he called the leverage cycle. Almost nobody read it. In October 2008 he sat in front of Ben Bernanke and the Board of Governors for 4 and a half hours explaining it, because by then it had eaten the world economy.

His argument contradicts a century of theory.

Everyone watches interest rates. The Fed sets interest rates. The textbooks solve for interest rates. Geanakoplos says the number that actually moves prices is the one nobody bothers to record: how much cash a lender makes you put down.

He went loan by loan to prove it. In 2000, homebuyers were putting 14% down. By 2006 they were putting down under 3%. Same houses, same people, 30 times the borrowing power. Prices doubled over those 6 years, and he argues that is where the doubling came from.

Two more numbers make the point. Banks buying AAA mortgage bonds had to post 1.6% of their own money, which is 60 to 1. Behind the entire 2.35 trillion dollars of toxic mortgage securities sat roughly 150 billion in real cash. In 2006, Bill Gates and Warren Buffett had about that much between them. 2 men could have owned the whole pile.

Here is what you can take from it.

Habit 1. Watch the down payment, not the rate. When lenders quietly cut what you have to put down, prices are about to rise. When they raise it, prices are about to fall, and nothing about the asset itself has changed.

Habit 2. Remember that a small group of optimists sets the price for everybody. Let them borrow 30 to 1 and a handful of buyers can clear an entire market. The price is their opinion, not yours.

Habit 3. Read the loan terms, not the headlines. In early 2007 the subprime index fell from 100 to 60 while actual losses were still under 1%. The stock market ignored it and peaked 9 months later.

Habit 4. Never let yourself go underwater. When a loan reached 160% of a home's value, 8% of those borrowers defaulted every single month. They had jobs. They just stopped paying for something no longer worth paying for.

He also had a fix, and he published it in the New York Times in 2008.

Take a family owing 160,000 on a house now worth 100,000. Cut the loan to 80,000. The family suddenly has equity, so they stay and they pay. The lender ends up with 80,000, which beats the 25,000 that a forced sale and 18 months of an empty house actually returns.

It never happened. The banks collecting those payments earned a percentage of the loan balance, so cutting the loan in half cut their own fee in half.

He was right in public, 3 years early, and nothing moved.

The Fed still does not track down payments.

[Kaien @kaienphase]

https://t.co/PbfztEwWL7

[Smoooothly @Smoooothly_]

INSTEAD OF SCROLLING FOR 1 MORE HOUR, SPEND THAT HOUR LEARNING WHY YOU KEEP LOSING MONEY TO YOURSELF.

Robert Shiller won the Nobel Prize in 2013. For a century, finance was built on the idea that people make rational money decisions. Shiller spent his career showing that they don't. He puts it more bluntly in his Yale lectures: people behave stupidly, and he includes himself.

He doesn't argue it. He runs a test on the students sitting in front of him.

He asks 200 of them to guess what the Earth weighs. Nobody knows, and nobody is expected to. Instead of a single number, each student writes down a lower bound and an upper bound, wide enough that they feel 90% certain the true weight sits somewhere between the two. A student who is honest about his own ignorance will write a very wide range.

Then Shiller reads out the real answer.

If those students judged themselves correctly, 9 out of 10 should have caught it. 1 out of 10 did.

They weren't wrong about the Earth. They were wrong about how much they knew about the Earth.

Guessing the Earth costs nothing. The same reflex applied to money is what Shiller spends the rest of the lecture on, and it shows up as 4 things you do without noticing.

You guard the small money and ignore the big money.

Lose $5 on the way to work and it ruins your morning, though you'll earn millions across your life. So you buy insurance on a funeral or a single flight, because those are small and visible, and you leave the loss that would actually wipe out your family uncovered.

You round every risk to zero or certain.

Your head holds 3 settings. Can't happen, maybe, will happen. A 2% chance of losing everything gets filed under can't happen, and then you never think about it again.

You let a random number set your price.

Kahneman and Tversky spun a wheel that landed on a number between 0 and 100, then asked people a question about the United Nations that had nothing to do with it. The higher the wheel landed, the higher people guessed. Afterward every one of them said it was a coincidence. You do this every time an asking price appears before you decide what something is worth to you.

You decide first, then go collecting proof.

Researchers phoned people who had just bought a car and asked which magazine ads they remembered. They remembered the ads for the car they bought and had skipped straight past the car they nearly bought. Once your money is in, you stop reading anything that argues with it.

Shiller's own uncle put his entire life savings into Ford stock, because Ford had employed him and treated him well for decades. Researchers sent actors with that exact portfolio to real financial advisors. 60% of the advisors told them to keep it. A client who hears that his loyalty is a risk walks out and never books a second meeting.

Nobody in that wheel experiment believed the wheel had touched them.

You can't delete the wiring. You can write your own number down before you look at the price.

[Kaien @kaienphase]

https://t.co/PbfztEwWL7