Hacker Newsnew | past | comments | ask | show | jobs | submit | willy_k's commentslogin

> If shit ever did hit the proverbial fan, you can bet that any US military walking around an American city would be constantly worried about getting a lead injection from any Tom, Dick or Harry who's got a firearm.

And then in many scenarios, that Tom, Dick, or Harry gets returned the favor, except with a tank round or worse. If the military remained organized, or at least large factions of it did, it would considerably outclass the general population in both intelligence and firepower / strike capabilities.

Soldiers would be on edge and vulnerable, sure, just like in the conflicts of the past few decades, but overall the military would retain a significant advantage.


> And then in many scenarios, that Tom, Dick, or Harry gets returned the favor, except with a tank round or worse. If the military remained organized, or at least large factions of it did, it would considerably outclass the general population in both intelligence and firepower / strike capabilities.

Tanks are obsolete weapons - esp in urban areas. They would be taken out by cheap drones. And you are forgetting that for every active US soldier there are ~3x retired and opinionated soldiers. Many of them know tactics to take down armor. And how to train civilians.


> Tanks are obsolete weapons - esp in urban areas. They would be taken out by cheap drones. And you are forgetting that for every active US soldier there are ~3x retired and opinionated soldiers. Many of them know tactics to take down armor. And how to train civilians.

Arguing over irrelevant details; all that can be true, and the military can therefore respond not by firing tank shells but by drone strikes, which many of the retired soldiers don't know how to respond to because almost all knowledge of modern drone warfare is in the Russian and Ukranian militaries right now.

In this scenario, if the US civilians were very lucky, the US military have learned nothing from the was Russian and Ukranian forces battle today; likely real scenarios are much messier, internal splits within both military and civilians on Trump/not Trump (different to Dem/Rep) lines, militia with ??? training, criminal gangs taking advantage, reduced international trade (perhaps except for whover supplies drone parts who may prop up both sides at the same time to maximise reveue, perhaps not because that's China and they want factories not rubble piles made out of factories), etc.


And I’m going to be using Valgrind in a few weeks, writing C in university now.


Ironically an AI written article.


The router seems to only apply to the ChatGPT version, not the API, so it’s not really anything new. Gemini already has functionally dynamic reasoning effort.


Yes they do, if the model size / vram requirement keeps shrinking for a given performance target, like has been happening, then it gets cheaper to run X level of model.


For API users, yes, but for the average person with a subscription or using the free tier it’s the inverse.


That doesn’t really matter for the point at hand, which is that Europe’s overregulation stifles innovation and international competition. Unless you’re arguing that the reason Europe hasn’t produced a SOTA LLM is because it’s not a net benefit, in which case that’s ignoring that it’s going to happen regardless and if Europe supposedly has its head on straighter than others, it would be most beneficial for the leading LLM to be European.


> in which case that’s ignoring that it’s going to happen regardless

we dont know the long term effects that llms have on a society that does not have these regulations in place. having europe as a backup in case the usa falls apart is probably a good idea


Regardless of what happens with LLMs, I predict that another country will leave the EU before a US state again attempts to secede from the union.


Replace "leading LLM" with progressively more evil things and see at what point the "if they exist, they better be European" logic stops seeming logical to you.


Innovation is not the same as improvement, but improvement without innovation is highly unlikely.


FWIW the fair comparison for Gemini 2.5 Pro would be o4-mini. That being said I’ve also found that Gemini is way better at getting it right on the first or second try and responding to corrections.


You don’t have to be happy with it. If Microsoft already has all your data it’s much easier to trust them for AI than to add another entity into the equation, especially OpenAI.


I’m as much a fan of thoughtless AI forwarding as the next HNer, but this seems like a fair situation for it:

https://chatgpt.com/share/68199e8e-5ad4-8012-8026-09fa353b60...


Chat gpt is known to make up citations and quotes. You shouldn't reference it unless you've checked every one for accuracy.


o3 is known to use web search to reference documents directly and be perhaps a bit overly diligent when it comes to double checking itself. Of course it can still make mistakes, but for something like this that’s a lot less likely than with 4o. Besides, I didn’t reference it, I just shared the results of a query as a third party to the thread (who decided to take as much time checking the claims as the two comments above me). Although it seems that the info is all coming from Wikipedia anyway so I might as well have shared that.


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: