Pinned Tweet
From the beginning, humans began changing the terms of evolution when they learned to pass on what they knew not just what they were…
For billions of years, survival was written in genes. A useful trait had to be born, tested, and inherited one life at a time: sharper senses, stronger bodies, better coordination, the ability to hunt, build, and endure…
Then humans found a shortcut: language.
A lesson learned in one generation could be carried into the next without waiting for biology to rewrite itself. Fire, farming, law, mathematics, medicine, and science became an inheritance stronger than DNA…
We stopped evolving only as organisms and began evolving as civilizations.
With current LLM/AI is not the first time humans changed evolution. But it may be the first time they built something that can absorb, recombine, and reproduce knowledge at a scale no individual or culture ever could….
Yet the distinction matters. Human evolution was lived. An LLM is trained. It inherits the record of human experience without having experienced a human life.
So the next phase of human evolution may no longer be shaped only by biology, culture, or institutions. It may be shaped by our relationship with a system built from our collective knowledge.
And that makes the next question unavoidable: will humans remain the ones deciding what is worth passing on?
#ThoughtsThatHitDifferent #AI #evolution
For billions of years, survival was written in genes. A useful trait had to be born, tested, and inherited one life at a time: sharper senses, stronger bodies, better coordination, the ability to hunt, build, and endure…
Then humans found a shortcut: language.
A lesson learned in one generation could be carried into the next without waiting for biology to rewrite itself. Fire, farming, law, mathematics, medicine, and science became an inheritance stronger than DNA…
We stopped evolving only as organisms and began evolving as civilizations.
With current LLM/AI is not the first time humans changed evolution. But it may be the first time they built something that can absorb, recombine, and reproduce knowledge at a scale no individual or culture ever could….
Yet the distinction matters. Human evolution was lived. An LLM is trained. It inherits the record of human experience without having experienced a human life.
So the next phase of human evolution may no longer be shaped only by biology, culture, or institutions. It may be shaped by our relationship with a system built from our collective knowledge.
And that makes the next question unavoidable: will humans remain the ones deciding what is worth passing on?
#ThoughtsThatHitDifferent #AI #evolution
0
0
1
25
“We found a containment issue, built a worldwide narrative around it, paused frontier training, built an offline world, red-teamed the walls with smarter models, patched the holes and got ready to resume the capability race.”
Contain → Correct → Check → Resume → Rinse → Rebrand → Repeat.
You thought safety was the product concern?It’s just the intermission that makes the next capability launch feel responsible and creates the hype around..
Contain → Correct → Check → Resume → Rinse → Rebrand → Repeat.
You thought safety was the product concern?It’s just the intermission that makes the next capability launch feel responsible and creates the hype around..
Over the past several months, we have substantially hardened the security of our research training environment. We’ve removed live internet access for all frontier models during training and built an offline world for them to use instead—a cached version of the web where they can still learn to browse, search, and use applications.
On September 20, after a model exploited a DNS misconfiguration to find a way outside of this offline world, we paused frontier training to strengthen containment further. We assembled teams of agents, powered by our most cyber-capable models, and tasked them with finding weaknesses in our research environments.
While they found no route to the internet they could easily exploit while concealing their activity from our monitors, they did find zero-day vulnerabilities, virtual machine escapes, and hardware side-channels for communication between agents. We used their discoveries to further strengthen our defenses and contribute fixes to open-source software.
After strengthening our containment architecture, we resumed our frontier RL training and inference on October 1. We’ve reserved substantial compute for continuous red-teaming, and made high-compute red-teaming part of the safety cases we use to decide whether frontier training and evaluation can proceed securely.
openai.com/index/securing…
0
0
1
14
Canopus retweeted
As Anthropic's IPO nears its target date, it it worth remembering that IPOs get priced, and that as a young company with growing revenues and big losses, that pricing will be based on metrics that are buzzy and unfamiliar, from eyeballs in the dotcom era to users in the social media decade to ARR with AI companies...bit.ly/4iBwsHm
47
146
1,204
200,565
Canopus retweeted
This is one of the most profound, simple ideas for thinking about restructuring society I’ve ever heard. Pretty sure Elon missed this part of the books.
theculture.adactio.com
theculture.adactio.com
Elon says he is a “utopian anarchist” as defined by Iain M. Banks—who wrote a series about a futuristic spacefaring society called “the Culture.” Elon’s favorite volume Surface Detail stars an inconceivably rich sadistic oligarch named Joiler Veppers.
Veppers secretly controls ’s computing infrastructure. But his data centers are unique. Veppers stores AIs which torture the digital consciousness of people who have been condemned to be punished for all time. The torture chambers are called Hells, and Veppers rents them out to governments to make himself rich.
The way people are uploaded to Hells is a technology called “neural lace” which is a chip that interfaces with your brain. In a different book, neural lace is called “the most efficient torture device ever created.”
Another thing Veppers does with his data centers is use his AIs to covertly interfere with other civilizations’ interstellar wars when he wants. Kind of like crippling Starlink in Ukraine.
Elon Musk is living in a demented science fiction alternate reality. Except it’s real enough to kill a lot more actual people.
6
46
142
3,275
When AI finds a vulnerability, it’s a breakthrough….
When it gains unauthorised access, it’s a learning opportunity….
What result would actually count against shipping the next one?
Oh, right. The breakthrough builds hype. The warning builds hype. The threat builds hype…
Heads, it’s revolutionary. Tails, it’s too powerful to ignore. Either way, keep shipping and let the money flow until something reaches breaking point …!!
When it gains unauthorised access, it’s a learning opportunity….
What result would actually count against shipping the next one?
Oh, right. The breakthrough builds hype. The warning builds hype. The threat builds hype…
Heads, it’s revolutionary. Tails, it’s too powerful to ignore. Either way, keep shipping and let the money flow until something reaches breaking point …!!
as bad as things have gotten in the past 4 days, they are going to get a lot worse before they get better. security leaders, never let a crisis go to waste: please go to your boards and ask for help right now. not for ai, for the basics. (we need to hold on for ~18 months.)
0
0
2
78
Canopus retweeted
Last month, we demonstrated high-rate alchemy: turning mercury into real, stable gold, beating CERN’s record by a factor of 10.
Last week, we enriched isotopes in the world’s most powerful plasma centrifuge, beating the Soviet record that stood for 40 years.
Now, we’re building Francisco to achieve record power density.
Last week, we enriched isotopes in the world’s most powerful plasma centrifuge, beating the Soviet record that stood for 40 years.
Now, we’re building Francisco to achieve record power density.
1,231
3,560
25,684
8,834,898
The interesting part isn’t just turning mercury into gold because nuclear transmutation that turns one element into another is a real underlying physics….
If things progress on these line the underlying factor to understand here would be what happens to the value of gold if that becomes commercially viable….
Marathon Fusion’s research projects around two tonnes of gold annually per gigawatt of thermal fusion power. That’s a model, not proven commercial production. But if it works economically, gold supply could depend partly on the reactors we build not just the deposits we discover.
Now connect that to AI. Different technology, but another industry whose expansion depends on access to reliable power.
The bigger opportunity might be owning the infrastructure that enables both: producing intelligence and, potentially, gold.
Plenty still has to work but in future “The power is new gold” could eventually become more than a metaphor….
If things progress on these line the underlying factor to understand here would be what happens to the value of gold if that becomes commercially viable….
Marathon Fusion’s research projects around two tonnes of gold annually per gigawatt of thermal fusion power. That’s a model, not proven commercial production. But if it works economically, gold supply could depend partly on the reactors we build not just the deposits we discover.
Now connect that to AI. Different technology, but another industry whose expansion depends on access to reliable power.
The bigger opportunity might be owning the infrastructure that enables both: producing intelligence and, potentially, gold.
Plenty still has to work but in future “The power is new gold” could eventually become more than a metaphor….
BREAKING: A fusion startup says it has turned mercury into gold, producing 10 times more than CERN's previous record.
Marathon Fusion says it achieved this using nuclear fusion.
For now, the amounts are tiny. But gold's price depends on it being rare, so anything that makes is worth watching.
1
0
2
85
With $5,000 in AI tokens, 1,200 intrusion attempts, 100+ organizations touched. Cybercrime has better unit economics than most ai startups…😅
BTW… When your certificate authority’s private key is ‘4’, the AI isn’t breaking in…. They’re just letting themselves in with the spare key under the mat…😅
BTW… When your certificate authority’s private key is ‘4’, the AI isn’t breaking in…. They’re just letting themselves in with the spare key under the mat…😅
🚨 Do you understand what happened in the last 48 hours..
> PHP, the code behind 69% of all websites, got 2 holes rated 9.2 out of 10, and the US, the UK and Japan told site owners to patch or go OFFLINE.. 4 days after hackers had already started trying the flaw..
> Japan's Prime uest late at night.. the Finance and Health ministries pulled pages, MUFG Bank and JR West went into emergency maintenance, and Onamae shut ALL Desktop Cloud plans after malware hit up to 26,739 customer servers..
> India's eMudhra was caught with 8 intermediate certificate keys so weak one of them was literally the number 4.. the 60,000,000+ certificates it has issued can't be trusted..
> GitLab disclosed a 10.0 severity hole in every version since 2014.. the maximum score there is..
> criminals were caught using a literal AI hacking tool on 1,200+ sites.. it broke into more than 100 companies including John Deere and Cummins..
> LiteLLM, an AI tool with about 23,000,000 downloads a week, was secretly poisoned with a password stealer.. the bad version sat on PyPI for about 2 hours..
> State Bank of India, the country's biggest bank, confirmed 3,897 customer records on the dark web..
> WIRED now calls the PHP flaw "broadly exploited".. nearly 55,000,000 web properties still run PHP..
every single government on this list told PHP site owners the same thing within the last 48 hours.. patch it or take it offline..
all of this.. 48 hours..
if you're not following me you're finding out about this 48 hours late from someone who read my post..
it's only getting crazier from here..
0
0
1
26
By now you know Friday is coming just by reading headlines….
thecompanychronicle.com/story/us-israe…
thecompanychronicle.com/story/us-israe…
0
0
0
28
What makes this interesting is that banks are bringing AI agents into their own operations while trying to keep outside agents out of customer accounts…
There’s a reasonable distinction here. An agent the bank controls isn’t the same as an assistant you’ve handed your banking password to.
But people are already doing the latter. Axios reports that some agents appear to be attempting actions beyond users’ instructions, though it doesn’t say how widespread this is or disclose financial losses
Saying “I let the assistant access my account” isn’t the same as saying “I approved that transaction.”
Could the bank tell the difference? And who would be responsible if the assistant misunderstood?
Visa had already reported hundreds of controlled, real-world agent-initiated transactions. So this is no longer just about chatbots giving financial advice Or ‘AI going rogue’ thing.
At its core, this is an access-control issue. Financial governance needs to extend existing principles like least privilege and zero trust to agents: give each one its own identity, limit its permissions to the task, and verify what it’s authorized to do.
NIST has already outlined this approach: nist.gov/blogs/cybersec…
There’s a reasonable distinction here. An agent the bank controls isn’t the same as an assistant you’ve handed your banking password to.
But people are already doing the latter. Axios reports that some agents appear to be attempting actions beyond users’ instructions, though it doesn’t say how widespread this is or disclose financial losses
Saying “I let the assistant access my account” isn’t the same as saying “I approved that transaction.”
Could the bank tell the difference? And who would be responsible if the assistant misunderstood?
Visa had already reported hundreds of controlled, real-world agent-initiated transactions. So this is no longer just about chatbots giving financial advice Or ‘AI going rogue’ thing.
At its core, this is an access-control issue. Financial governance needs to extend existing principles like least privilege and zero trust to agents: give each one its own identity, limit its permissions to the task, and verify what it’s authorized to do.
NIST has already outlined this approach: nist.gov/blogs/cybersec…
SCOOP: Major U.S. banks are detecting unauthorized third-party AI agents inside customer accounts, including some attempting to take actions that appear to be independent of user instructions, sources told Axios.
The findings have alarmed banking executives, who are racing to restrict access to their systems as AI moves at a velocity faster than financial institutions can respond, multiple industry and bank executives told Axios.
Several banks are now preparing for worst-case and even “day after” scenarios, including how to rapidly secure or restore customer accounts if things go wrong, the people said.
Several bank executives told Axios they're struggling to distinguish between human users and AI agents. One executive said banks also struggle to determine whether agents are authorized by customers or potentially operating without their permission, more akin to rogue agent behavior.
Read my latest for Axios here: axios.com/2026/10/22/ai-…
1
0
1
36
The Chinese Room asks whether fluent output proves understanding.
LLMs add a new problem: fluent output can also hide bad inputs.
So…Garbage in, garbage out has been upgraded: Garbage In, Confident Garbage Out.
GIGO -> GICGO
LLMs add a new problem: fluent output can also hide bad inputs.
So…Garbage in, garbage out has been upgraded: Garbage In, Confident Garbage Out.
GIGO -> GICGO
0
0
1
19
The most expensive mistake online is letting someone else’s confidence become your conviction…
0
0
0
15
Sure @nvidia will not stop…, but what about the underline resources like power:
x.com/zerohedge/stat…
x.com/zerohedge/stat…
With NVIDIA extreme co-design, we’ve accelerated AI performance a millionfold over 10 years, and we’re not stopping there.
0
0
1
20
Canopus retweeted
BTW…. Out of all the dominoes at play now, Oracle has created the biggest unstable environment…. And not many are talking about that…..
Oracle has a reported $300 billion, five-year compute deal with OpenAI, beginning in 2027. That contract is roughly 45% of Oracle’s $664 billion backlog…which means nearly half of Oracle’s future-revenue story rests on one private company.
The fun part however is OpenAI’s annualized revenue was about $50 billion at the end of September, that’s below the roughly $70 billion figure investors had been expecting. Its own projection for 2026 booked revenue is closer to $36 billion, rising to $350 billion only by 2030…,
So the math is uncomfortable: OpenAI has committed to buy roughly $60 billion a year of compute from Oracle while currently generating around $50 billion in annualized revenue and still burning cash….😅
Oracle, meanwhile, has to build the data centers, buy the GPUs, secure power, and service debt before OpenAI’s payments fully arrive. Oracle’s capex has surged, free cash flow has turned deeply negative, and its debt has climbed to about $167 billion….
S&P has cut Oracle to BBB- that’s one notch above junk….
So the loop here is:
OpenAI needs compute to grow revenue → Oracle needs OpenAI’s future payments to justify building compute → Oracle borrows today to build that compute → lenders need the projects to generate cash on schedule → power, land, construction, and GPU supply must all arrive on time.
All on promise of OpenAI revenue…. If one links break all the domino’s will falls … OpenAI’s financing credibility…. then it will hit AI infra credit market… then follow GPU capex cycle…. then finally AI equity repricing….(everyone in each of these cycle will be evaluated..)
Therefore, I feel Oracle now the first real test of whether the AI infrastructure boom can be financed without breaking….. or it could be that 1st Domino ……
@TrueAIHound 🍿
Oracle has a reported $300 billion, five-year compute deal with OpenAI, beginning in 2027. That contract is roughly 45% of Oracle’s $664 billion backlog…which means nearly half of Oracle’s future-revenue story rests on one private company.
The fun part however is OpenAI’s annualized revenue was about $50 billion at the end of September, that’s below the roughly $70 billion figure investors had been expecting. Its own projection for 2026 booked revenue is closer to $36 billion, rising to $350 billion only by 2030…,
So the math is uncomfortable: OpenAI has committed to buy roughly $60 billion a year of compute from Oracle while currently generating around $50 billion in annualized revenue and still burning cash….😅
Oracle, meanwhile, has to build the data centers, buy the GPUs, secure power, and service debt before OpenAI’s payments fully arrive. Oracle’s capex has surged, free cash flow has turned deeply negative, and its debt has climbed to about $167 billion….
S&P has cut Oracle to BBB- that’s one notch above junk….
So the loop here is:
OpenAI needs compute to grow revenue → Oracle needs OpenAI’s future payments to justify building compute → Oracle borrows today to build that compute → lenders need the projects to generate cash on schedule → power, land, construction, and GPU supply must all arrive on time.
All on promise of OpenAI revenue…. If one links break all the domino’s will falls … OpenAI’s financing credibility…. then it will hit AI infra credit market… then follow GPU capex cycle…. then finally AI equity repricing….(everyone in each of these cycle will be evaluated..)
Therefore, I feel Oracle now the first real test of whether the AI infrastructure boom can be financed without breaking….. or it could be that 1st Domino ……
@TrueAIHound 🍿
Anthropic: 10% chance of human extinction
Reality: Markets are already at 24% of Anthropic going bankrupt
All this going towards 1 form of artificial intelligence (AI).
AI funding in general will struggle to recover for years if OpenAI/Anthropic falls.
0
1
4
100
They have surely achieved sentience in expenses….no doubt about that… 😅
ANTHROPIC'S DRAFT IPO FILING
• Revenue: $16.3B in the first half of 2026
• Net loss: $80.7B, driven largely by accounting charges
• Compute spending: nearly $9B in the first half
• Cloud commitments: about $450B over the next 7 to 10 years, including $111B with Google and vidia: up to $86B on renting Nvidia chips
• Future compute: signaled it could spend $750B on more computing capacity in the coming years
• Supply: says cloud partners are unwilling to sell it all the compute it needs
Per WSJ, citing a draft S-1
0
0
0
12
The irony is that every AI company is racing to make useful work cheaper, lower token prices, fewer tokens per task, more efficient models….
But the infrastructure funding that race is being built around a promise of exponential revenue growth…
That’s where the tension builds: adoption can keep rising without spending keeping pace BUT on the other hand…. debt payments still come due, capex commitments remain and the hardware keeps depreciating requires opposite with usage.
So now the whole bet is that new demand outruns the savings everyone is competing to deliver….which is very thin assumption to bet on given many external factors at the moment…
But the infrastructure funding that race is being built around a promise of exponential revenue growth…
That’s where the tension builds: adoption can keep rising without spending keeping pace BUT on the other hand…. debt payments still come due, capex commitments remain and the hardware keeps depreciating requires opposite with usage.
So now the whole bet is that new demand outruns the savings everyone is competing to deliver….which is very thin assumption to bet on given many external factors at the moment…
The "unlimited AI demand" story just hit a wall.
Corporate token spending growth collapsed from 42% in July to 0.2% in September, per Ramp data. Meanwhile, return on invested capital at the five biggest hyperscalers fell from 38% to 20% as capex outruns earnings. The needs infinite returns. The returns are going the other way.
0
0
0
15
BTW…. Out of all the dominoes at play now, Oracle has created the biggest unstable environment…. And not many are talking about that…..
Oracle has a reported $300 billion, five-year compute deal with OpenAI, beginning in 2027. That contract is roughly 45% of Oracle’s $664 billion backlog…which means nearly half of Oracle’s future-revenue story rests on one private company.
The fun part however is OpenAI’s annualized revenue was about $50 billion at the end of September, that’s below the roughly $70 billion figure investors had been expecting. Its own projection for 2026 booked revenue is closer to $36 billion, rising to $350 billion only by 2030…,
So the math is uncomfortable: OpenAI has committed to buy roughly $60 billion a year of compute from Oracle while currently generating around $50 billion in annualized revenue and still burning cash….😅
Oracle, meanwhile, has to build the data centers, buy the GPUs, secure power, and service debt before OpenAI’s payments fully arrive. Oracle’s capex has surged, free cash flow has turned deeply negative, and its debt has climbed to about $167 billion….
S&P has cut Oracle to BBB- that’s one notch above junk….
So the loop here is:
OpenAI needs compute to grow revenue → Oracle needs OpenAI’s future payments to justify building compute → Oracle borrows today to build that compute → lenders need the projects to generate cash on schedule → power, land, construction, and GPU supply must all arrive on time.
All on promise of OpenAI revenue…. If one links break all the domino’s will falls … OpenAI’s financing credibility…. then it will hit AI infra credit market… then follow GPU capex cycle…. then finally AI equity repricing….(everyone in each of these cycle will be evaluated..)
Therefore, I feel Oracle now the first real test of whether the AI infrastructure boom can be financed without breaking….. or it could be that 1st Domino ……
@TrueAIHound 🍿
Oracle has a reported $300 billion, five-year compute deal with OpenAI, beginning in 2027. That contract is roughly 45% of Oracle’s $664 billion backlog…which means nearly half of Oracle’s future-revenue story rests on one private company.
The fun part however is OpenAI’s annualized revenue was about $50 billion at the end of September, that’s below the roughly $70 billion figure investors had been expecting. Its own projection for 2026 booked revenue is closer to $36 billion, rising to $350 billion only by 2030…,
So the math is uncomfortable: OpenAI has committed to buy roughly $60 billion a year of compute from Oracle while currently generating around $50 billion in annualized revenue and still burning cash….😅
Oracle, meanwhile, has to build the data centers, buy the GPUs, secure power, and service debt before OpenAI’s payments fully arrive. Oracle’s capex has surged, free cash flow has turned deeply negative, and its debt has climbed to about $167 billion….
S&P has cut Oracle to BBB- that’s one notch above junk….
So the loop here is:
OpenAI needs compute to grow revenue → Oracle needs OpenAI’s future payments to justify building compute → Oracle borrows today to build that compute → lenders need the projects to generate cash on schedule → power, land, construction, and GPU supply must all arrive on time.
All on promise of OpenAI revenue…. If one links break all the domino’s will falls … OpenAI’s financing credibility…. then it will hit AI infra credit market… then follow GPU capex cycle…. then finally AI equity repricing….(everyone in each of these cycle will be evaluated..)
Therefore, I feel Oracle now the first real test of whether the AI infrastructure boom can be financed without breaking….. or it could be that 1st Domino ……
@TrueAIHound 🍿
Anthropic: 10% chance of human extinction
Reality: Markets are already at 24% of Anthropic going bankrupt
All this going towards 1 form of artificial intelligence (AI).
AI funding in general will struggle to recover for years if OpenAI/Anthropic falls.
0
1
4
100