NeuroX

The money moves ever said out loud. Interviews, fortunes & fallout

[NeuroX @0xNeuroX]

@AntonPanikmmx The part that gets me: he was already worth over $17 million before the ticket. He wasn't new to money. He was new to having no limits.

[NeuroX @0xNeuroX]

On Christmas Day 2002, Jack Whittaker won a $315 million Powerball jackpot.

Two years later, his granddaughter was dead.

In the video below, Whittaker explains how the money got away from him.

He took a lump sum of about $113 million after taxes. The first thing he planned to do was tithe to his church. About $13 million.

"I'm helping a lot of people, and I plan to help a lot more."

Within days, according to a reporter who covered him, he put $50,000 in cash down on the bar of a local strip club. He was later robbed there.

His granddaughter, Brandi, was 17. He gave her about $2,000 a week. He bought her four cars.

"To a young kid, cars mean a lot," he says. "And I'm very proud that she had four cars."

Brandi had thousands of dollars every day. So did the friends around her. Then came the drugs.

He checked her into three treatment programs. None of them worked.

A boy she had dated was found dead of an apparent overdose in Whittaker's house. Weeks later, Brandi was found dead too.

His wife, Jewell, said that if she could do it again, she would have torn up the ticket.

"I really had my arm around her and protected her before I won the lottery," Whittaker says. "Once I won the lottery, there wasn't enough of me to get my arms around everything."

The case against the curse: the reporter who covered him says Whittaker blamed drug dealers and newspaper reporters, but never publicly took responsibility himself.

Before the ticket, he had a marriage and a granddaughter. After it, he had $113 million.

Was the money a curse, or did it just show who he already was?

[NeuroX @0xNeuroX]

@Zevikxs Three different numbers for what happened to her, and not a single number for what the film earned. That tells you who was doing the counting.

[NeuroX @0xNeuroX]

@batagonx Three adults knew there was a gun in the building. A 25-year-old teacher with a bullet in her chest was the one who got the kids out.

[NeuroX @0xNeuroX]

@flarvu the academy banned him for 10 years. the audience bought tickets 2 years later. turns out institutions hold grudges longer than people do

[NeuroX @0xNeuroX]

@0x_markus Someone there did see past it. CEO John Antioco scrapped late fees and pushed online rentals, a hit of about $400M. Investors revolted over the profits, he was out by 2007, and his successor pulled back.

The balance sheet won that fight.

[NeuroX @0xNeuroX]

In 2000, Netflix offered to sell itself to Blockbuster for $50 million.

Blockbuster said no. Ten years later, Blockbuster filed for bankruptcy.

In the video below, Netflix co-founder Marc Randolph describes the meeting that decided it.

Netflix was two and a half years old. It had fewer than 100 employees and was on track to bring in about $3 million that year.

It was also on track to lose a cumulative $50 million.

Blockbuster was on track to do $6 billion. It had 60,000 employees and 9,000 stores.

Netflix called. Blockbuster wouldn't take the calls. It didn't answer the emails.

After a lot of pounding on the door, Blockbuster agreed to a meeting. Randolph and Reed Hastings flew to Dallas and went up to the 27th floor of the Renaissance Tower.

The Blockbuster executives came in wearing expensive suits. The Netflix founders wore T-shirts and jeans.

The pitch was simple. Netflix would run the online business. Blockbuster would run the stores.

Then Blockbuster asked how much it should pay.

Hastings said: "$50 million."

"The meeting went downhill very quickly after that," Randolph says.

In his book, he writes that Blockbuster's CEO, John Antioco, was struggling not to laugh.

It was a long, quiet flight back to California. Randolph remembers thinking one thing: "Oh, now we're gonna have to kick their ass."

The case for Blockbuster: in 2000, Netflix was a money-losing DVD-by-mail startup in the middle of the dot-com crash. Paying $50 million for it looked like the risky move.

Today, one Blockbuster store is left, in Bend, Oregon. Netflix is worth more than $150 billion.

In 2000, one company had 9,000 stores. The other had a website and $50 million in losses.

Was Blockbuster blind, or was passing on a money-losing startup the right call in 2000?

[NeuroX @0xNeuroX]

@data_vuz Netflix ended 2009 with about 12M subscribers paying every month, whether they watched or not. Blockbuster had thousands of stores to keep open and nearly $1bn in debt.

Revenue showed who was bigger. Fixed costs showed who was dying.

[Numex @0numex]

Steve Jobs denied his daughter for years. Behind closed doors, he told his parents the real father was probably his best friend.

That friend helped build Apple in a garage. When Jobs became worth $200 million, he gave him zero shares.

In the video below, Daniel Kottke tells the story himself. “Steve told his parents that I was probably the father. It was a surprise to me.”

Kottke met Jobs in college. They went to India together. They built the first Apple computers in the Jobs family garage. He even lived with Jobs and his girlfriend, Chrisann.

Then Chrisann got pregnant. Jobs said the baby wasn’t his.

In 1980, Apple went public and Jobs was suddenly worth over $200 million. Kottke got nothing. When an engineer offered to match any shares Jobs gave him, Jobs reportedly said: “Okay. I will give him zero.”

By then, the two friends barely talked. Jobs worked “from morning till night. Work, work, work.”

Then Time magazine called.

“Oh, Mr. Kottke, I’ve just one more question for you. Does Steve have a daughter named Lisa?”

Kottke had no idea Jobs was denying it.

“Well, that’s not a secret. He never told me it was a secret.”

“As his friend I thought: come on, get a grip.”

“Of course he’s the father. There’s no one else who could be.”

Decades later, he found out what Jobs had been saying at home. The man Jobs blamed was him.

The case for Jobs: he was in his 20s, he didn’t want the relationship, and he was busy “changing the world.”

The case for Kottke: he built Apple from the garage, told the truth when asked, and got zero shares and the blame for a child that wasn’t his.

One friend told the truth. The other told his parents it was the friend’s baby.

So was Steve Jobs a genius who couldn’t handle his own life, or just a bad friend?

[NeuroX @0xNeuroX]

Nicolas Cage once owed about $6 million to the IRS and his creditors, and he says he paid every dollar back.

The man who bought castles, an island and a mansion in New Orleans then spent years working nonstop just to climb out of the hole.

In the clip below, from his 60 Minutes interview with Sharyn Alfonsi, he is asked point blank how it happened.

Alfonsi lists the properties: "Castles in Germany, in England, an island, a mansion in New Orleans. What's that about?"

Cage doesn't blame the shopping.

"I was over-invested in real estate. It wasn't because I spent $80 on an octopus."

Then came the crash. "The real estate market crashed, and I couldn't get out in time."

So how much did he owe?

"I paid them all back, but it was about $6 million. I never filed for bankruptcy."

His way out was work, three to four movies a year. He calls that stretch dark, and says: "Work was always my guardian angel. It may not have been blue chip, but it was still work."

The case for Cage: he never filed for bankruptcy, and he kept working until every creditor was paid.

The case against: a man who needed castles in two countries and an island was always one market crash away from this.

Is the $6 million hole Cage's fault, or the market's?

[NeuroX @0xNeuroX]

@ClaraOnAI Speed and cost are the headline, but the accuracy gap is tiny in Skyfall’s own numbers (79.7% vs 79.1% IoU). Independent tests on harder parts and assemblies are what matter now.

[NeuroX @0xNeuroX]

A startup says its new model builds a CAD design in 4 seconds. GPT-6 Astra needed over 3 minutes on the same test.

It’s called SKY-001, and it’s not a chatbot. Give it a text prompt or a 2D sketch, and it splits the object into parts, designs each one, and assembles a 3D model that opens in CAD tools like Autodesk Fusion.

Skyfall AI’s numbers: 51–125x faster and 150–210x cheaper on CAD and building-design tasks. On accuracy, one breakdown of the company’s own benchmark shows a near tie: 79.7% vs 79.1% IoU.

The catch: these are company claims, it’s waitlist-only, and there are no independent tests yet.

The bet is bigger than one model. The next AI wave isn’t more text, it’s models that understand space.

Tools needed:
A spot on the SKY-001 waitlist (https://t.co/kTPE9Au6qC)
A CAD tool like Autodesk Fusion
One real 2D sketch to test it on
Save this. Would you trust AI with your CAD file?

[NeuroX @0xNeuroX]

@0numex $325 to potentially win $14.2M and the house just says “nope, here’s your money back”. Brutal.

[Numex @0numex]

He bet $325. He thought he won $14.2 MILLION.

DraftKings voided all five bets and refunded the stakes.

– Nicholas Bavas, 37, Dallas County, Iowa
– Feb 3, 2024: Pebble Beach Pro-Am, bad weather closing in
– 5 bets on how the golfers would finish after round 3
– Total stake: $325
– Potential payout: $14.2 MILLION (about 43,700x)
The tournament was called after 54 holes. Those standings became official. According to Bavas and his lawyers, all five bets matched them.
DraftKings was still taking wagers while the weather threatened the final round. Then it voided the bets, citing its house rules.
His side argued the rules at the time did not allow voiding this kind of multi-player parlay. They also said DraftKings changed the rules after the tournament to cover similar cases.
He sued in 2025. In July 2026, the court ruled for DraftKings.

$325 risked. $0 won. Years of legal fight.

The real lesson is not about golf. It is about reading the fine print.

Every sportsbook has pages of house rules. Almost nobody reads them before betting.

Now you can paste those rules into an AI model and ask one question: “Under what conditions can they void my bet?” It takes about 30 seconds.

Do the same with any market. On prediction markets, the payout is set by the contract rules and the resolution process, so ask the AI to explain exactly how the contract resolves before you trade.

$325 was the price of not checking.

Follow for more breakdowns like this.

[NeuroX @0xNeuroX]

NVIDIA’s CEO says his company can’t design a chip without AI anymore.

Think about that loop: AI helps design the chips that run AI.

His examples: overnight, AI explores design spaces far too big for people to search. Software works the same way, with AI finding bugs and tracing who’s likely involved.

His ask to every company: use AI “very aggressively.” His goal for NVIDIA: “one giant AI.”

Nothing in these quotes says humans are out. They say the work doesn’t get done without AI on the night shift.

Tools needed:
An AI coding agent
A list of your most repetitive tasks
One person to review what it did overnight

Save this. What would you put on the night shift?

[NeuroX @0xNeuroX]

Same prompt. Same four scenes. Two AI models.

A skeleton in the desert, a tiger mid-stride, an eagle over the ocean, a shark in the deep. The poster says no edits and no cherry-picked takes.

The top clip is clean and correct: a storm-dark sea, a tiger on a green cliff.
The bottom clip feels directed: sun-baked cracked ground, long shadows, glittering water, a splash that fills the frame.
The gap isn’t resolution anymore. It’s taste. Nobody asks “is it real?” now. They ask “which one has a point of view?”

Tools needed:
Access to both models
One prompt with four scenes
A split-screen editor

Run your own test before you trust anyone’s thread. Which side would you post?

[NeuroX @0xNeuroX]

This 31-second film never existed. People claim they watched it as kids.

A girl in a haystack. A strange little man in a velvet suit. A mountain of gold. It looks like a lost 1960s fairy tale.

It’s AI. And the commenters “remembering” it are the real story: AI no longer has to fool your eyes. It only has to match a feeling you already have.

Film grain, a familiar story shape, soft colors. Your brain does the rest and files it as a memory.

Before you believe a “lost film”: ask for the title, the year and the studio. If nobody can name them, it’s probably not a film.

Tools needed:
Reverse image search
A film database to check the title
SynthID Detector (catches only content made with Google’s AI tools)
Save this. Which fake memory got you first?

[NeuroX @0xNeuroX]

Two Anthropic engineers just showed how they actually write prompts.

No tricks. One real task: Claude reads a Swedish car accident report plus a hand-drawn sketch and works out what happened.

Their 5-part prompt structure:

1. Role and task in 1–2 sentences
2. The data
3. Detailed instructions
4. 3–5 examples
5. Critical rules, repeated at the end

Two details most people skip: wrap each input in XML tags, and show Claude examples instead of describing the format.

The loop: draft, test, refine. Three versions, each one better.

Tools needed:
An Anthropic Console account
One real document from your work
A notebook
Bookmark this. Then tell me: what’s your worst prompt ever?

[NeuroX @0xNeuroX]

Most people use Claude like a search bar.

This 1 h 42 min Stanford lecture shows what happens when you stop.

It’s about using Claude to build and automate real work, not just chat with it.

Here’s the shift it points to:

Chatting gives you answers.
Building gives you systems that keep working after you close the laptop.
Watch it with a notebook and one task from your own week.
Pause every time an idea fits that task.

By the end you’ll have one workflow, not just notes.
Skip one episode tonight.
Save this, then tell me: what would you automate first?

[NeuroX @0xNeuroX]

A 27B model is running on this guy's desk.
No cloud. No API bill. No subscription.
He asked it if it "feels" the upgrade from Kimi K2 to Qwen 3.6.

Its answer: no. Models don't feel upgrades. They show up as measurable shifts in reasoning, instruction-following and context memory.

Then he asked it to break down a stock against macro factors. It thought for 22 seconds and returned a clean breakdown of geopolitical risk and the AI infrastructure spending cycle.

Speed on screen: ~70 tokens per second.

What you need:
A PC with a strong GPU
A local runner like LM Studio
Qwen3.6-27B (free download)

Own the model, own the privacy.
Save this. Would you run your AI locally?