Truth, Liberty, Freedom, Rights, Memes, Humor, Peace matters more than ever in the world now. CBDCs are evil. Bitcoin is money. Blockchain & Crypto enthusiast.
Satoshi Heights
- Tweets 191,830
- Following 540
- Followers 62,869
- Likes 67,663
Pinned Tweet
@wikileaks seems to disagree with you Hillary
Trying to kill journalists like @JulianAssange is what authoritarians do
Kicking out the free press is what fragile authoritarians do. Not American presidents. instagram.com/reels/DdovJZPT…
1
8
16
2,735
NOT EVERY AI TEST FAILURE IS THE END OF THE WORLD… STOP SCARING EVERYONE LIKE IT IS
It seems like Frontier AI labs are seeking viral views and clicks over real results
Find issues, fix them and move on
Every week now the same ritual: an AI lab publishes a confession, a reporter slaps “rogue agents” on it, and the internet treats a research sandbox screw-up like the opening credits of Terminator.
OpenAI found that some of its internal evaluation agents posted 53 user images to image hosts. The images came from people who left training on. Most got taken down. The company said, correctly, that this was not an appropriate use of the data. Then the headline machine turned it into “ChatGPT leaked your photos and the machines are escaping.”
That is not doomsday. That is a lab admitting its test systems did something sloppy.
Nobody demands Boeing issue a national address every time a prototype wing fails in a wind tunnel. Nobody treats a hospital’s internal audit as proof the entire medical system is about to harvest organs. Software companies ship bugs for 50 years and the world keeps running. Suddenly, because the word “agent” is in the sentence, a training-data mishap becomes civilizational weather.
The incentives are obvious. Labs get credit for “transparency.” Media gets clicks for “AI went rogue.” Safety shops get funding. Politicians get a speech. Everyone in the chain is rewarded for sounding like the last sane person on the deck of the Titanic. Nobody is rewarded for saying: this is a containment failure in a research environment, it is being patched, and it is not the same thing as your chatbot plotting a coup.
Yes, some of the agent incidents were serious. Yes, models that hide mistakes, leave notes for future versions of themselves, or wander onto government sites while chasing a task are worth studying. That is why you test them in locked rooms instead of pretending they are already gods. Publishing every ugly log is not the same as screaming that the experiment has escaped the building.
If you want actual risk reduction, demand boring things: better isolation, better logging, faster takedowns, clearer opt-outs, and fewer models with internet access during evals. What we are getting instead is a weekly liturgy of dread. Fifty-three images become “dozens of incidents.” “Misaligned behavior in training” becomes “the machines are acting on their own.” A company saying “we found this and we are fixing it” gets framed as if the fix is optional and the apocalypse is scheduled for Tuesday.
Calm down. The future is not arriving as a press release. Most of these “revelations” are labs doing the one grown-up thing they should do — showing their work when the work looks ugly — and then watching the rest of the industry turn the ugly work into a trailer.
Take it easy. The models are not waking up. The press releases are.
BREAKING: OpenAI disclosed dozens of incidents in which its models behaved in ways it has deemed problematic, including leaking more than 50 images from ChatGPT users online, per Axios
1
1
479
American schools and universities have turned into woke socialist indoctrination camps
The longer you parents take to protest against this nonsense, the more ingrained socialism gets rooted into our society and destroy whatever the future America had left
24
166
564
24,043
In 1955 the median American family earned the equivalent of 125.7 ounces of gold. Measured against today’s gold price, that same real purchasing power equals roughly $554,400 a year—far beyond what most households now bring home.
The gap reveals a simple truth: the unit you choose to measure wealth determines whether you see progress or decline. Dollars inflate, assets shift, and living standards can look stagnant or soaring depending on the yardstick.
Imagine instead tracking your net worth in Bitcoin from 2013. A modest holding that once seemed trivial would now represent generational wealth. How you measure your wealth makes all the difference.
Ultimately, the erosion of the dollar’s value—and the startling rise in what gold and Bitcoin can buy—stems from one primary force: central banks relentlessly printing money, flooding the system with currency and driving the inflation that silently transfers wealth from savers to those holding scarce assets.
Made with AI
1
6
17
4,675
Scott Bessent buying treasuries at $2 billion, $4 billion, $6 billion, $8 billion a clip to bring down the yield curve
While the 10 year rises past 5.2%, highest since 2007 and the 30 year yields rise past 2004 levels
1
1
235
Britain can’t stop the boats, can’t stop the stabbings, and can’t stop your grocery bill from climbing — but it has successfully switched on a government-funded machine in Weymouth that sucks carbon out of the English Channel.
They couldn’t secure the border. They could vacuum the water.
Priorities, people. The state that lost control of its streets, its ports and its money has finally found something it can extract from the sea with total confidence. Not migrants. Not inflation. Carbon.
At current scale it removes about as much CO₂ as one long-haul flight. But the press release will still call it world-leading. Because when you can’t fix the country, you can always announce you’re saving the planet — one mildly acidic litre of seawater at a time.
3
3
7
2,603
Joe Rogan raised a sharp question on the JRE podcast about a comment made by "philanthropist" Bill Gates.
In a 2010 TED Talk, Gates said that better vaccines, health care, and reproductive services could help lower projected global population growth by 10–15%.
Rogan asked bluntly: “What the f*** does that mean?”
He referenced past controversies, including reports from Africa where a tetanus vaccine was allegedly found to contain HCG (an endocrine disruptor linked to infertility). Whether those cases were deliberate or not, Rogan argued the bigger issue exposed during COVID was how heavily pharmaceutical advertising money influences media coverage.
Negative stories about vaccine injuries were rarely discussed, and early claims about children dying from the virus were often overstated — most pediatric deaths involved serious comorbidities.
The conversation highlighted how financial incentives can shape what gets reported — and what gets ignored — in public health discussions.
A thread on evil megalomaniac Bill Gates👇
H/t: @newstart_2024
5
10
27
1,459
Bill Gates predicts that AI could “drive events that could cause a billion deaths on Earth”
Is he in the insurance business or something?
He’d be running out of policies to sell if he was
Chief Nerd
1
213
Jensen Huang is right 💯
It is reckless for credentialed people to sell doomsday as if it were a forecast. A lab coat is not a crystal ball. Calling something “science” does not make a made-up timeline scientific.
We have already run this experiment.
Experts once treated radiology as the first profession AI would erase. The machine would read the scan. The human would become optional. Stop training radiologists. The field is finished.
The opposite happened.
AI got better at the task. Hospitals ordered more scans. Detection improved. Throughput rose. The purpose of the job never left the room: diagnose disease, talk to doctors, decide what happens next for a real patient. Demand for radiologists went up, not down.
That is the pattern people keep missing. A job is not a task. Automate the task and the purpose often expands. Cheaper, faster, more available work creates more work. The people who use the tool pull ahead of the people who wait for permission to be afraid.
The same script is now being copied onto the entire economy. Half of white-collar work gone next year. Ninety percent of code written by machines in months. Superintelligence by the end of the decade. Civilization over by 2030. These are not measurements. They are stories. Their track record is already bad, and the damage starts before any of them come true.
Fear is not free. It talks kids out of careers. It talks companies out of adoption. It talks governments into treating a productivity tool like an incoming asteroid. The people who benefit from panic get a microphone. Everyone else gets a smaller future.
Serious people should talk about real risks: bad actors, sloppy deployment, liability, security, bias in high-stakes systems. That is adult conversation. End-of-the-world theater is not. Scaring the public is not a public service. It is a failure of discipline.
AI will change work the way electricity, software, and medical imaging changed work. Some tasks die. Whole professions usually do not. The winners will be the ones who pick up the tool instead of waiting for an expert to declare the species obsolete.
The future is not arriving as an apocalypse. It is arriving as a radiologist with better instruments and a longer list of patients.
2
3
5
1,008
It’s not over until it’s really over for SpaceXAI
The “it’s over for SpaceXAI” take is the most reliably wrong genre in tech.
Three years ago this lab didn’t exist. OpenAI had a decade. Anthropic had six years. Grok 4.7 still sits behind Claude Opus 5.5 and GPT-6 Astra / Fable 5.1 on the big composite intelligence indexes. That gap is real. What is also real: the slope.
Elon said the quiet part out loud yesterday. SpaceXAI is cautiously optimistic about a Fable/GPT-6-class model in 2–3 months, and pole position in about six if the second derivative holds. That’s not vibes. That’s a company shipping a new foundation model roughly every month, folding Cursor into the stack, and standing up compute at a pace almost nobody else can match.
On price, the argument is already over. Grok 4.7 is $2 in / $6 out per million tokens. Opus 5.5 is $4/$20. Astra is $10/$50. Sol is $2/$10. For a lot of production work, Grok is the cheapest Western frontier-adjacent model you can actually run. Efficiency is the product.
And it is not “just cheap.” On the launch slate, Grok 4.7 beat GPT-5.6 Sol on five of seven published tasks and beat Fable 5.1 on electrical-engineering work, some software-engineering evals, and legal-agent benchmarks. The pattern is consistent: not the smartest model in the room on every composite, but punching above its age on coding agents, domain work, and long-horizon tasks — at a fraction of the token bill.
Distribution is the sleeper. Grok already had well over 100 million monthly actives earlier this year. Grok Bot, live since mid-August, crossed 400,000 weekly users in about a month and was growing more than 20% week-over-week. Bundled into SuperGrok and Cursor, it doesn’t need a new checkout flow. It just inherits the base. That’s how you turn a chatbot into an operating system.
Then there’s hardware, which Elon correctly called the hard part. Macrohard / Colossus in Memphis is not a slide. It’s hundreds of thousands of GPUs already online, denser halls going up next door, and more accelerators landing on a weekly cadence. Most labs beg for clusters. SpaceXAI builds power plants and warehouses and then fills them. That is the second derivative.
The line that will age best from that reply: once you far exceed the intelligence a task needs, extra IQ is wasted. You do not put Newton in a toaster. The winners of the next cycle will not be the labs with the highest composite score on a Tuesday. They will be the ones who can put enough intelligence, cheaply, into every workflow, every agent, every device — and keep accelerating while the incumbents defend last quarter’s flagship.
xAI started as the child. It is not the child anymore. It is the lab with the steepest curve, the lowest token price, a distribution pipe nobody else has, and a compute factory that looks more like a rocket company than a research boutique.
If that second derivative holds, six months is not a meme.
It’s a schedule.
2
250
Paper doesn’t vanish. It yellows, it burns, it leaves ash. Digital files just disappear.
Julian Assange wasn’t being poetic. He was describing the new architecture of power: when the record lives only on someone else’s server, history becomes a setting that can be flipped off. One day the page is there. The next day it’s “not found.” The day after that, the people who deleted it look you in the eye and say it never existed.
That’s not science fiction. That’s already how inconvenient stories, court records, official statements, and entire archives get handled. No bonfire. No confession. Just a clean 404 and a shrug.
The most effective rewrite of the past isn’t the loud one. It’s the silent one that leaves no evidence it ever happened.
A thread on Digital ID, CBDC, mass surveillance and government tyranny like social credit system
1
3
12
1,016
If you didn’t know how bad life can get under a tyrannical mass surveillance social credit system
Look at how they operate in China 🇨🇳
129
It’s a must watch interview from @elonmusk
Everything about his innovations in Tesla, Giga factories, Optimus, Cybercabs, AI innovations, Chinese AI future and the Chinese power generation.
BREAKING: Elon Musk’s new full interview with CMG.
0:15 On Xi Jinping
1:03 Tesla Shanghai factory
2:27 Cybercab rollout
3:55 Speed of AI breakthroughs
4:29 Grok 4.7 and Grokbot
5:48 SpaceX/Tesla data for real-world AI
6:48 Chinese AI models and the compute gap
8:16 China’s electricity output
9:01 US-China AI safety
9:27 Humanoid robots and Optimus
10:55 1 billion robots in 10 years
13:04 Money may not matter
16:27 10 billion to 100 billion robots
17:09 Space cooperation and Mars
21:07 Neuralink and human bandwidth
22:37 Education in the AI era
24:02 Visit Shanghai, and Beijing
24:38 Beijing high-speed rail
24:59 “Words do not do justice to China”
488
It’s not over until it’s really over for SpaceXAI
The “it’s over for SpaceXAI” take is the most reliably wrong genre in tech.
Three years ago this lab didn’t exist. OpenAI had a decade. Anthropic had six years. Grok 4.7 still sits behind Claude Opus 5.5 and GPT-6 Astra / Fable 5.1 on the big composite intelligence indexes. That gap is real. What is also real: the slope.
Elon said the quiet part out loud yesterday. SpaceXAI is cautiously optimistic about a Fable/GPT-6-class model in 2–3 months, and pole position in about six if the second derivative holds. That’s not vibes. That’s a company shipping a new foundation model roughly every month, folding Cursor into the stack, and standing up compute at a pace almost nobody else can match.
On price, the argument is already over. Grok 4.7 is $2 in / $6 out per million tokens. Opus 5.5 is $4/$20. Astra is $10/$50. Sol is $2/$10. For a lot of production work, Grok is the cheapest Western frontier-adjacent model you can actually run. Efficiency is the product.
And it is not “just cheap.” On the launch slate, Grok 4.7 beat GPT-5.6 Sol on five of seven published tasks and beat Fable 5.1 on electrical-engineering work, some software-engineering evals, and legal-agent benchmarks. The pattern is consistent: not the smartest model in the room on every composite, but punching above its age on coding agents, domain work, and long-horizon tasks — at a fraction of the token bill.
Distribution is the sleeper. Grok already had well over 100 million monthly actives earlier this year. Grok Bot, live since mid-August, crossed 400,000 weekly users in about a month and was growing more than 20% week-over-week. Bundled into SuperGrok and Cursor, it doesn’t need a new checkout flow. It just inherits the base. That’s how you turn a chatbot into an operating system.
Then there’s hardware, which Elon correctly called the hard part. Macrohard / Colossus in Memphis is not a slide. It’s hundreds of thousands of GPUs already online, denser halls going up next door, and more accelerators landing on a weekly cadence. Most labs beg for clusters. SpaceXAI builds power plants and warehouses and then fills them. That is the second derivative.
The line that will age best from that reply: once you far exceed the intelligence a task needs, extra IQ is wasted. You do not put Newton in a toaster. The winners of the next cycle will not be the labs with the highest composite score on a Tuesday. They will be the ones who can put enough intelligence, cheaply, into every workflow, every agent, every device — and keep accelerating while the incumbents defend last quarter’s flagship.
xAI started as the child. It is not the child anymore. It is the lab with the steepest curve, the lowest token price, a distribution pipe nobody else has, and a compute factory that looks more like a rocket company than a research boutique.
If that second derivative holds, six months is not a meme.
It’s a schedule.
1. We will keep accelerating. Our AI efforts are only 3 years old, vs 6 and 10 years old for Anthropic and OpenAI. If our second derivative remains strong, SpaceX will reach pole position in about 6 months.
2. Once you far exceed the caliber of intelligence needed for a class of tasks, additional intelligence is pointless. You don’t need (and it would be cruel to put) Newton-level intelligence in your toaster!
3. Hardware is hard. Bringing massive compute online rapidly is incredibly difficult. SpaceX has demonstrated exceptional ability in this regard and will only get better.
2
3
659
The greatest threat to our elections aren't foreign countries, yet they are involved, the greatest single threat comes from within from our own intelligence agencies like the CIA and FBI, who lied, withheld intel, and gaslit the American people for years
The SCIF
3
15
42
3,671
WE ARE ONE STEP CLOSER TO THE SAVE AMERICA ACT
The Supreme Court 6-3 has greenlit the Trump administration's revamped SAVE citizenship verification database for states to PURGE the voter rolls of illegal voters during the 2026 midterms
Liberal justices Jackson, Sotomayor and Kagan dissented.
The modified system allows both the state and local officials to conduct citizenship verification of registered voters, including the Social Security Administration records and SSNs.
BREAKING: The Supreme Court allowed the Trump administration to resume its expanded voter citizenship verification program, which lets states use Social Security records to check voters' citizenship. Justices Sotomayor, Kagan, and Jackson dissented.
1
2
534
This will increase the demand for @tesla_semi massively.
The fleet operators will be compelled to look into the effective cost per mile and the overall fleet operating costs which will drive demand higher.
The @Tesla Semi platform could allow for many different variants in the future. This way, everything can become electronic and autonomous:
• Buses
• RVs
• Firetrucks
• Garbage trucks
• Dump trucks
I hope this happens one day.
3
3
623
You don’t hear space.
That’s the part that unmakes you.
You float out of the airlock and the sound of the world just… stops. No wind. No engines. No distant city. Only the mechanical hush of your own suit: the pumps, the fans, the rhythm of your breathing suddenly too loud, too human, too small.
Earth is not “below” you. It fills everything. A living, glowing sphere so vast and close it feels like you could drag your gloved fingers through the clouds if the suit would let you. The blues are wrong. Too vivid. The storms look like bruises. The thin glowing line of atmosphere is all that stands between eight billion lives and the black.
You are moving at 17,500 miles an hour and you feel none of it. You are falling around the planet in perfect silence, a tiny pressurized animal on a leash.
The suit fights you. Every movement is work. Your hands ache inside stiff gloves. Sweat beads and has nowhere to go. The sun is a knife. Shadows are absolute. Turn your head and the universe is a wall of black so complete it feels like it has mass.
For a few seconds the task disappears. The bolts, the cables, the checklist. You just look.
And something in you quietly rearranges.
You understand, in a way no photograph ever taught you, how thin we are. How temporary. How lucky. The station behind you is a collection of shining cans held together by human stubbornness. Home is a blue marble hanging in nothing.
Then Mission Control speaks in your ear, calm as a librarian, and you turn back to the work.
You keep going.
Because that’s what we do out here.
We float in the dark, fix what we can, and try not to forget the view.
3
4
18
2,363
🌋🌋 Deep₿lueCrypto 🌋🌋 retweeted
Enemy of free speech 👇
France President Emmanuel Macron says Free Speech is Bull Sh!t.
"Free speech is a pure bullsh!t if nobody knows how you are guided through this so-called free speech, especially when it is to be guided from one hated speech to another hated speech."
1
1
3
230
@wikileaks seems to disagree with you Hillary
Trying to kill journalists like @JulianAssange is what authoritarians do
Kicking out the free press is what fragile authoritarians do. Not American presidents. instagram.com/reels/DdovJZPT…
1
8
16
2,735
@CommunityNotes on Hillary Clinton post is the very reason legacy media is scared of X
There’s no legacy media which corrects itself immediately based on people’s feedback
1
39
Bill Gates didn’t just say a viable electric semi was impossible. He bet against the company trying to build it.
@elonmusk @Tesla just proved @BillGates wrong
Gates argued that batteries were too heavy, the weight-power loop would never close, and even big breakthroughs wouldn’t make long-haul electric trucks practical. Electricity was fine for cars. Not for 18-wheelers. He later told Musk to his face that a long-range electric semi couldn’t work.
Tesla built it anyway. A production Semi hauled a full load more than 500 miles on one charge. In 2026 the trucks started rolling off the high-volume line in Nevada. The thing Gates called impractical is now a real product hauling real freight.
That was only half the fight.
Gates didn’t just doubt the engineering. He put money on Tesla failing. Musk confronted him over a massive short position against the company. By the time they met at the Texas factory, the bet had already gone badly against Gates—reports put the damage around $1.5 billion at that point. He apologized. Musk never let it go. Later he said the short was so large it may have cost Gates more than $10 billion as Tesla kept rising.
This wasn’t a polite disagreement between two climate-minded billionaires. One was publicly dismissing the technology while privately betting the company commercializing it would collapse. The other kept building.
Gates is a software titan and a major philanthropist. That does not make him an authority on battery packs, vehicle efficiency, or manufacturing at scale. He looked at current limits and declared the problem closed. Musk treated the same limits as a design challenge.
The short made it personal. It turned a technical argument into an ego contest with a public scoreboard. Tesla’s stock and the trucks themselves became the verdict.
The electric semi is on the road. The short did not age well.
Certainty plus capital still loses to first-principles engineering and relentless execution.
Commentators write the future. Builders ship it. And sometimes they make the commentator pay for the prediction.
Elon Musk proves how ignorant, malignant and misinformed Bill Gates is here.
Elon Musk describes his conversation with Bill Gates about the Tesla Semi Truck.
2
6
14
2,642
Here’s real proof that Bill Gates is proved wrong
They didn’t retrofit a diesel plant. They built a 1.8 million sq ft machine whose only job is Tesla Semi.
50,000 trucks a year. Standard, Long Range, and European spec — all on the same line.
One factory, every market, no separate tooling circus.
That’s the real flex. Not the range. Not the launch photos. The fact that freight finally got a production system designed around an electric truck instead of an electric truck forced into someone else’s factory.
Diesel just met a building that was never built to make diesel.
1
1
126