A lot of Anthropic’s moves make sense if you follow the LessWrong / rationalist community writings on AI safety. A lot of it is distilled in Ant’s blogs and leadership interviews and podcasts (Amanda Askell is particularly interesting).
Ant’s models, culture and leadership actions are largely consistent with their beliefs, even if they may seem flawed / incomprehensible.
Relevant anecdote: I interviewed with them for a MTS role in 2023. I think the technical part went fine but the interviewer was clearly frustrated by my low regard for LLM safety. I didn’t get the role.
I'm assuming this is popular because of Fable restrictions. AFAIK, open source is not excluded from ITAR / EAR restrictions (or other export restriction in other countries).
So the real solution you're looking for is technology that can't be arbitrarily gatekept by a sovereign nation.
I’ve been exceptionally displeased with Claude Code since end of February and switched completely to Codex in April. The blasé way in which one person (Borris) capriciously changes the system prompt multiple times a day, also no longer writing his own prompts (whatever that means).
That, the 5 different secret levers you have to pull to make it not stupid, the fact you hs e to go to the guy’s twitter account to find all the un-dumbing features and flags that aren’t documented anywhere else. That they decrease thinking budgets silently when they run out of compute instead of announcing the rationing, and gaslighting users at every step of discovery. The fact that internally they have their own coding harness and don’t use Claude Code primarily. The lack of formal evals and consideration for millions of users collective hundreds of millions of hours of investment in their workflows — that’s all off the top of my head, let me tell you how I really feel about what they did to Claude Code..
I adore gpt5.5 and maintain my own codex fork - but I have no idea how long I’ll get this performance / cost - I know it won’t be forever. I’d like to know precisely how much it’ll cost in hardware to run a gpt5.5 open source model locally. Hell a lifetime license to a model I can run locally is also be open to.
But I like building my own tools, from software to physical shop tools. I like being able to rely on my tools.
More responding here to the assertion that this is blowing up due to Fable.
> We saw some of this behavior in the last administration too. US voters deserve better.
With due respect, this take is very deluded. US voters have very little to lose if the tech is not available to the rest of the world. US politicians and elite, regardless of political inclination, understand the enormous strategic potential of this technology and will ITAR the shit out of frontier models and/or use them as leverage for extracting concessions out of other countries.
The main losers are Big AI labs, their investors, foreign employees and rest of the world.
Fwiw, China and other countries would’ve done the exact same thing. It’s perhaps the game theoretic optimal approach when your comparative advantage is so vast (capital, compute, talent, embedded knowledge) and keeps growing especially if RSI is real (making it nearly impossible for anyone else to catch up)
>With due respect, this take is very deluded. US voters have very little to lose if the tech is not available to the rest of the world.
Really? You think the economics of the AI buildout remain viable if US companies cannot export their highest value services?
You think expelling foreign AI researchers doesn't hurt the industry or boost foreign competitors? Half (or whatever) of Google's AI team, including their AI chief are foreign nationals and/or located outside of the US.
You think that other IT exports will not suffer if the US turns out to be an unreliable and even capricious supplier?
I humbly accept it's very difficult to game this out with any degree of confidence, especially as other countries deploy more resources.
But the questions about viability of the labs because of export restrictions is not in my cone of uncertainty. If you believe the labs' implied/stated objectives, the end goal is eating all human-driven GDP, and the US is still the largest single-market economy in the world, last I checked. Keeping the politics of AI-driven unemployment aside, economy-wide automation would make the US wealthy beyond imagination.
US exports as a whole is only 10% of the GDP. I am not aware of the international revenue share for OpenAI/Anthropic.
>If you believe the labs' implied/stated objectives, the end goal is eating all human-driven GDP, and the US is still the largest single-market economy in the world, last I checked.
The US is about a quarter of the global economy, but let's use Microsoft's international revenue share of ~50% as a proxy for tech services. It's ~40% for the S&P 500.
I don't know what share of that would be impacted by export bans, but it would certainly affect ROI. It would hurt the competitiveness of the wider US tech industry and create incentives for moving highly paid work overseas.
>US exports as a whole is only 10% of the GDP.
It's 12% to 13% but this is distorted by the way in which tech services are counted in export statistics and also by tax avoidance. Just look at Ireland's ridiculous GDP numbers.
If the technology is as powerful as these somehwat fantastical "goals" suggest, the incentive to use it everywhere in the world would be enormous. An export ban wouldn't mean that only the US has the capability. The theory behind current models is well known. It's just a matter of optimising them for specific use cases and using them on an industrial scale.
Most AI researchers are not US citizens either. It's completely obvious that this is the US shooting itself in the foot (if it were to last, which I don't believe).
For other sovereign nations, being a market for American technology products is no longer good enough to get access to American tech. I imagine the relationship would become a lot more one-sided (in favor of America) if the model capability gap keeps increasing.
To the rest of the world, US aggression is perceived as oppressive and unilateral primarily because it’s a political act meant to further domestic political goals. All US administrations act in their own self interest, as they should.
But the US being the world’s cop was never going to be sustainable indefinitely. It was partly because of hegemonic power and partly because of a reluctance of wealthy democratic countries to seriously invest in their military when they could just “pay” someone else.
I don’t think anyone wants the pre world war 2 order. So we are likely going to see otherwise reluctant states step up militarily to protect their neighborhoods.
Brakes are being added to actuators but they're more useful for static holds / locking than dynamic balancing. Standing even in humans is a dynamic balancing activity.
It is, but you don't have to do it at every joint all the time. If you lock your knees your front and back thigh muscles don't have to work on balancing in that axis at all.
You evaluate the merits of the content? From a high-level systems pov, the article is largely correct, even if some details might be missed / simplified
How do you evaluate the merits of a graph of information? I usually read articles like this to learn something, not to grade someone else’s assignment.
As a lay person, I have limited knowledge of this field, yet am extremely interested.
Unfortunately AI gen content is being used widely to spread spurious information/fake news etc. So my knee jerk reaction to AI gen content is - "this is going to be fake".
If the information you are trying to convey is true, and is technical/objective in nature, then why shy away from associating your identify to it?
I feel the same until I’m reminded I’m paying Anthropic $100 every month for something that’s indispensable to me now and would probably pay a lot more. Very inelastic demand as long as competition is low at the frontier.
Why will Anthropic be able to do that? That assumes there won't be competition and switching costs will be high. That doesn't seem to be the case to me.
There are no places in the US that Tmobile is the only wireless mobile network provider. While all 3 mobile network providers have weak coverage areas, Verizon is considered to have the most reach.
I think about this a lot - on HN and elsewhere online, there's an outsized portion of people who are likely to see use in AI. If you're online, you're on a computer and you're likely to also do office work that an AI can help with / do for you.
I don't think a majority of people will find it actually useful in the long run. I do game development as a hobby outside of work and every artist I know is outright hostile towards any sort of AI, to the point that they are dropping out of any community that so much as allows any mention of AI workflows.
I know that OpenAI is at least exploring "AI as entertainment" (Sora) but it has yet to be seen if that will be widely accepted or profitable. I've also been reading about teens talking to chatbots more and more rather than talking to real people, which seems like it will only end in mental health disasters.
$100/month isn't much for developer tooling. If you add up how much I spend on hardware upgrades, other SaaS products like backup services, software licenses, and other things it's easy to justify $100/month for a powerful tool.
I pay for my own AI provider subscriptions because keeping work and personal strictly separated is important for me. I do know some people who secretly pay $200/month for Claude and use it at their job even though it's not approved. I do not recommend doing that, but it shows that some people value this for their work.
For developers earning more than $10K per month, spending less than 1% of salary on tooling to make the job easier is easy to justify.
I’ve been a copilot and ChatGPT subscriber for probably close to two years now, give or take a couple of months, and I had a trusted friend telling me for months to give Claude a try.
It took about two weeks of really running it through its paces, and constantly slamming against the limit on it to convince me I had to upgrade to at least the 100/month sub, and at this point I wouldn’t blink to bump that to the 200/month if necessary.
I 100% believe we’re in a bubble, and that this level of compute isn’t sustainable at this price point, but for as long as I have it, I’m going to run it at the redline.
I’m a solo dev working on a project that I’ve just gone full-time on, after about 1.5 years of part time work. It’s a codebase that I laid the groundwork in, and has very well established systems, standards, and constraints.
The work I’m using Claude to do is the exact work I would be doing myself, but it does it at somewhere in the neighborhood of 5-10x the pace I could have. I don’t know that I could get the same rate of production if I managed a team of 2-3 programmers. Right now, it’s literally almost perfect at taking my iterative suggestions, and implementing them at that accelerated pace.
Honestly the hardest part is dealing with the fact that at the end of the day, I have to understand this codebase perfectly (solo dev and all that), so I have to take in changes to it that are also 5-10x the rate my normal intuition would. But, again, the plus side is that it’s implementing them essentially exactly as I would have, as it has ~20k lines of code that I wrote to use as an example.
If I were to hire even one other programmer, I’d be paying well north of 5k/month, and I’d not only be managing a super computer programmer tool, but an actual human being as well. $100/month might as well be free comparatively.
Because I don’t believe 50% gross margins at face value, as being discussed in this thread I think the economics of all of these things are far more complex than that.
For what it’s worth, I haven’t staked my investment portfolio on there being a bubble, I’m just preparing for the worst and doing as much work as I can with Claude before a potential massive rug pull happens.
>If you’re using it for personal work, why is $100 worth it?
I'm not who you were replying to, but:
My work pays for $100/mo Claude, I pay another $100 to bring it up to $200/mo level because:
- Partly: I got in the habit back when work was only paying $20 and I was paying the $180.
- It is not worth it to me to spend braincells trying to optimize my use to slip into the $100 plan, I give everything "Opus, effort max" and with the $200/mo plan I never run out ($100 I'll run out mid-morning).
- I run a *lot* of experiments, including work-related and personal, to try to understand and improve my AI use skills.
- I also use it for a lot of personal things, right now I'm using it to help me plan a backyard studio and ADU.
Agreed on both, especially hallucination. That's what makes their chat app even worse, it's very opaque about web search and sources, so you can't tell whether it's a hallucination.
Acetaminophen (paracetamol) is the drug of first choice for addressing pain and fever, in India at least. To the extent that it's regularly abused, and I know people who have been hospitalized because of abuse.
Even then, doctors are usually disapproving of ibuprofen (or some combination of it with paracetamol) unless paracetamol is contraindicated for some reason, and I had always wondered why.
I did listen to this 99% Invisible story about the use of NSAIDs in India once[1]
What you describe in an interesting contrast to the situation in The Netherlands.
Here, virtually no one is prescribing ibuprofen _without_ also prescribing a baseline of paracetamol.
Ant’s models, culture and leadership actions are largely consistent with their beliefs, even if they may seem flawed / incomprehensible.
Relevant anecdote: I interviewed with them for a MTS role in 2023. I think the technical part went fine but the interviewer was clearly frustrated by my low regard for LLM safety. I didn’t get the role.
reply