I think this fundamentally is not how most nation states and powerful entities work.
It's hard enough for France or Russia to keep the lights on, let alone plot world domination. Government are essentially incompetent. There is no deep state. There are no Jason Bournes. There are no adults in the room. The smartest people in the world go into the private sector or academia while the CIA ends up with midwits like Elissa Slotkin. Someone here posted that obviously the US government must have a more powerful model than Anthropic. I find that take delusional.
And the Europeans? It's the year 2026, and they are still worried about climate change. They are fundamentally unserious people.
If you want to know where things are going, ask the people who actually know - researchers at Anthropic and OpenAI. They will tell you.
Yeah, it's really just about exposure.
Maybe the normie is using an equivalent of 5.6 Luna on light thinking mode. And the coder is using GPT 6 Astra on high thinking. And the AI researcher is using Aeon on ultra.
Now, in 3-6 months, everyone moves up one step. The future is already here, it's just not evenly distributed.
How do you know Claude Tax is getting it right,
This is the principal agent problem. Without being an expert yourself, how do you know your accountant is getting it right? They probably aren't. Every rich person I know is looking for an accountant because the one they have sucks.
People are already using Claude to complete 1000+ page returns, hilariously printing them out and mailing them to the IRS. The tech will never be worse than it is right now.
It's not there yet. Next year it could be.
But I assume that, had we achieved superintelligence, they still use lawyers for political cover. In the future, I think it's likely that white collar workers will have a fake job where AI does all the work and they put their "checked by a human" stamp on it, but in reality don't actually check.
Why aren't nation states doing anything?
I think you answered your own question.
Realistically for most people it just doesn't matter whether they're "prepared" or not. Either AI is a normal technology and people will figure out how it slots into their lives over time as it diffuses into the world and into their lives as needed, or it's the singularity and 99.999% of people will have absolutely no way to affect the outcome regardless of how much they monitor the situation.
Let's say I believe there's a good chance superintelligence happens in the next 18 months. What do I do with this information? What does anyone do with it? Do I live my best life now since the world's going to end? Do I work super hard to accumulate capital to avoid the permanent underclass? Do I just live my life since AI will be super wise and do the best thing for me?
I have no clue what to do. And neither do nation states. Why in the world would France ready its nuclear arsenal because of superintelligence? What possible purpose would that serve? There is no clear and obvious action to take, because the post-singularity future is inherently unpredictable.
Here's some things I actually would recommend:
- Don't die
- Don't get medical interventions that are risky and will likely get better soon (e.g. plastic surgery)
- Invest at least some part of your money in things that are positively exposed to AI (easy enough)
Other than that, what is there to do?
One bull case for Anthropic following IPO is that they will likely be able to run inference at much cheaper levels per unit of intelligence. Using their own AI models, which competitors don't have access to, they will be able to train smaller models that far outcompete Chinese models at a lower price of inference.
They will be able to serve Fable-class models at wildly profitable prices even as they lower the price.
Why do you think that he doesn't matter?
Let's say we forget about the visionary leadership or whatever. I think that's a mistake, but for the time being cast it aside.
Let's focus on the fact that Nvidia is utterly dependent on TSMC with whom Jensen has a close and very personal relationship. Do you think Morris Chang is going to give random white lady the same deals as he gave Jensen, or will he just shrug and give Apple or Google or a million other companies their allocation? Heck he might even knock them down a peg out of vengeance for perceived anti-Asian racism. Apparently its all handshake deals.
Who would you rate as a serious thinker then? They seem to me to have defined the age.
Yes, Elon is wildly optimistic. His stretch goals should not be interpreted literally. As Michael Scott once said "Shoot for the stars and if you fall at least you'll land on Mars".
The problem with FSD is that it is AGI complete because it has to read and react to various things that a human might encounter. It probably fails in many of the same ways that a recent immigrant from Punjab with a degree from a crooked driving school would fail - only much less catastrophically.
On the other hand, it is legitimately superhuman at doing things like paying attention and breaking in response to unsafe conditions. It might not be better than the best drivers but it's better than mean driver (heavily influenced by the left tail that is responsible for 80% of the fatalities).
I won't pretend it's perfect, but if everyone used it 95% of the time like I do I believe we'd have a lot fewer accidents.
All that said, the Cybercab is real, is it not??
Weird because I'm at 95% self-driving on my Tesla and Cybercabs are roaming the streets of Austin without steering wheels and drawing rave reviews.
Personally my timeline for FSD was 2030 but I think we'd easily beat that timeline if governments get out of the way.
I only wish that more drivers had FSD so that the roads would be safer.
Yes, I am a business owner so I understand the pain all too well.
Damn, my memory betrayed me. Nevertheless a remarkable turnaround for him since 2025 when he posted this:
"Why I don’t think AGI is right around the corner": https://www.dwarkesh.com/p/timelines-june-2025
Next time I suppose I will do a better effort post since there were a lot of good points that Dwarkesh made and it's clear that almost no one actually read the article. So it would have been better if I spent an hour or so to summarize it.
The board was feeling kinda dead though so I typed it out in 15 minutes.
How boned is Nvidia? Apparently Jensen had 60 direct reports and a handshake agreement with TSMC.
Too bad the HR ladies fired him.
Could see this stock falling by 50% within a few months if they don't bring him back.
If you put $1 million into TQQQ in 2010 and simply held it, you'd now have $288 million.
Leverage can be amazing.
Probably if you are young and high income it's worth using a little leverage, although its not currently a great time with 6% margin rates.
That was kind of tongue and cheek.
I agree with Dwarkesh that there's a 75% chance we will have an intelligence explosion involving RSI in 2027. That doesn't mean infinite intelligence of course, but it will mean that AI exceeds human experts in most fields. Hard to pin down an exact prediction for what that would involve, partially because AI is already much smarter than humans at things like math and people shrug. I'd guess that all white collar work will be replaceable, but much of it already is and momentum kind of keeps things going.
Has any AI skeptic ever said "I was fundamentally wrong"? Because it seems that they are always wrong and just obfuscate, obfuscate, obfuscate.
What are some things you are 100% positive that an AI won't be able to do in 2027?
You've never hired an accountant have you? They mess up all the time. It's actually comical how bad the median accountant is. I do my own taxes now for this reason. Claude tax, whenever it comes, will be a massive improvement in every way.
Possible use for humans post 2027: Claude tax might refuse some of the sketchier things that a client might want to do, whereas a human accountant might not.
Yes. Even if frontier AI stopped all today, we could replace nearly all accountants and financial planners with the technology we currently have.
So why hasn't it happened yet? It's Very simple to explain, actually. Frontier AI labs have limited resources and its best to spend those resources on more AI research. Why spend $$$ developing a tax product today when you can spend $ to build a better tax product in one year. So best to just pick enough low hanging fruit to pay the bills while spending almost all energy on further research.
This is similar to how, if we sent a probe to Alpha Centauri with today's tech, it would be passed by a probe that left in 10 years with better technology. (This is apparently called the "incentive trap" but it needs a better name).
The singularity is coming in 2027 and no one cares.
Dwarkesh is out with a new essay in which he argues that there is a high chance (not 100%, but like 75%) that we get an intelligence explosion in the next 12 months equal in magnitude to the jump between GPT-3.5 and GPT-6.
This jump would be caused by the automation of AI research.
I highly, highly recommend giving this a read or a listen. Dwarkesh is, in my mind, the most thoughtful public intellectual around today (sorry Scott) and is known for his in depth research and thoughtful interviews. In his essay, it's likely that he addresses whatever objections that you have that just sprung to mind, so just give it a read. Really.
At this point, it seems the only thing that would stop superintelligence from forming is an unknown unknown, as in there is some fundamental constraint or that hasn't been identified, or an emergent phenomenon that only reveals itself in the future. I guess this has all been rehashed a million times, and I don't have a lot to add beyond what Dwarkesh said, so I'll revert to the purpose of this board which is to make some social commentary.
People are completely unprepared for what's coming.
In financial forums that I visit, and in youtube videos I watch, people are entirely dismissive of what is happening at OpenAI and Anthropic. "It doesn't really think! They're losing money! Drug development, no way! It only works for software! The software is buggy! They'll be bankrupt in a year! Switching costs are zero! Cheap Chinese models will win!".
Etc.. Etc... Etc...
Some concepts are just too high IQ for most people. Probably most people with an IQ below 100 don't really understand supply and demand for example. And the floor for understanding recursive self improvement is a lot higher.
Everything feels pretty normal now. But there's at some point going to be a big oh shit moment when social consensus hits and 90% of the population goes from denialism to doomerism in a short period. I have a hard time predicting what will happen then except that I'd assign a 50% chance that Anthropic will be the most valuable company on Earth in 12 months. Cold comfort. Superintelligence is coming and it's going to change everything. Maybe it will be good...
Supposedly China's "one weird trick" was distilling OpenAI/Anthropic. That's been turned off and we haven't heard as much about the Chinese models these last few months.
It's the kind of thing only a moron would blame on Trump, but I'm sure they'll try.
He does seem to be losing the mandate of heaven though.
The only scenario where attacking is easier is if the vulnerability cannot be fixed.
This might be the key. The vulnerability that cannot be fixed is us. A human in the loop will always make cyber defense incomplete.
No comments on the proposal which I assume is tongue and cheek, but I'll take the opportunity to soapbox.
People, especially liberals, are legitimately deranged about women's competitiveness in sports. There are some people who think men and women have equal athletic ability, which is obviously wrong. But there are other people who think the difference is small, which is also wrong.
Anyone who has played co-ed sports knows that almost all the men are better than almost all the women, and when that's not true it's mostly due to differences in practice time or age.
But on the surface this isn't obvious. When we look at measures of raw athleticism, the difference isn't always huge. For example, at all distances, the difference between elite male and female running speed is only about 10%, perhaps 2 standard deviations.
But sports are not just about a single measure of athleticism, they are about compound measures.
So men might be +2std in speed, +2 std in strength per body weight, +2 std in height, +3 std in throwing ability, +1 std in spacial reasoning, +1std in decision making under pressure, etc...
With all the averages compounded, in a sport like American football, a 20th percentile man is equivalent to a 99.9th percentile woman given equal training time.
So basically on Thursday last week all of tech crashed on AI panic, probably because of data center bans.
Since then there has been a huge rebound. Except for Nvidia which declined another 9%.
So even though Nvidia is "only" down 9%, the other names in its sector are up. In the counterfactual with no Jensen scandals, Nvidia is likely trading 20-30% higher than now. It's a big delta an deservedly so.
The nightmare scenario here is Jensen starts a new company, raises $100 billion or so, poaches all of the top talent at Nvidia (remember 90% of employees wanted him to stay) and then uses his existing handshake relationships with TSMC to get allocations at the expense of Nvidia. While this new company might not immediately take market share, it would absolutely gut Nvidia. And probably employees would love the challenge, and the stock options.
The semicondutor industry cannot afford to be run by shopminders. Look at Intel for a stark example.
Although Nvidia might rally in the short term, I would use any rally as an opportunity to exit any positions. This one is dead money, IMO.
Jensen has been on the rocks for a week or so, thus the 5 day chart. But why do you think AMD and AVGO are up so much? Do you trade stocks often? What’s your favorite play right now?

It means that Anthropic does have a moat. For any level of intelligence, the frontier labs can provide it at the cheapest rate. They aren't going to be outcompeted by some random Chinese copycat.
If Anthropic isn't profitable, it's because they choose to invest into research. At any time, they could turn on the faucet and make 90% net margins.
Now obviously they won't do that because they still have competition from OpenAI/Google. But with 140 billion in their bank account, another 100 billion coming via the IPO, and a business that could be profitable if it wants to be, they are in an incredibly advantageous position. It's similar to how Amazon always chose growth over profit.
More options
Context Copy link