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

If the real cost where really 1900 / 50% = 3800 USD per month then you could rent your own H200 with 140 GB VRAM.

The issue currently is capacity. Servicing the models cost just peanuts of electricity but the popularity is killing them.


So enormously that the haven't had any sane permission system built in this year.

ESG is just giving up returns for a good feeling, isn't it?

Stock picking is just folly for individual investors, isn't it?

Anyone claiming they can consistently beat any large index is just delusional, aren't they?


On ESG/SRI: fair, excluding sectors comes at a cost, and we make that trade-off knowingly.

On stock picking: the system is rules-based and mechanical, not discretionary. The "folly" argument applies most strongly to human judgment calls, which this attempts to remove. I literally wanted to reduce bias and get a better vantage point.

On beating the index: 14 years of backtested data with walk-forward validation suggest it's possible for this specific strategy. Whether it holds going forward, nobody knows. We publish the ten best and worst precisely because we're not claiming certainty.


As Matt Levine often points out, there are two possible cases for ESG

1/ This will bring worse returns, but I'm willing to accept the loss in order to forward values I support

2/ This will bring better returns, since the market underrates risks from bad ESG companies (e.g. the long-term return on capital for coal companies will be worse than the market expects)

People marketing ESG funds (or anti-ESG, same rule applies) usually emphasise the second.

> Anyone claiming they can consistently beat any large index is just delusional, aren't they?

This is obviously not true. RenTech would like a word.


I recently looked at this a bit but came away with the impression that at least on API pricing the models should be very profitable considering primarily the electricity cost.

Subscriptions and free plans are the thing that can easily burn money.


The physical buildouts and massive R+D spending is the big part.

This! And with AI studio you get a couple of free calls per day (it has gotten less and less). I have had days where I would be able to get 100 USD worth of tokens from AI studio for free. 1m tokens in and great code out.

You can even turn most of the censorship off in the AI studio (but not the hidden top_k of 64 they force in there).

AI studio is where you go if you want an actually good mostly uncensored model. Gemini 3.1 is fully and somehow still quietly coomer approved.


The article is certainly firebranding, but the core tenet strikes a valid point: how has the US lost the plot within such a short time? How did it go from the flag bearer of freedom and progress to isolationist bully that wants to invade Greenland and become best friends with Russia?

From the outside it is really hard to comprehend. Was it FoxNews that poisoned the American mind or the social media brainwashing? How can a society allow a billionaire to cut programs in Africa that saved hundreds of thousands of lives that cost pennies when compared to any military adventures.


It seems like the US never really reovered fully from the Civil War, and the undercurrent of racism has just been allowed to fully come to the surface with social media.

> how has the US lost the plot within such a short time? How did it go from the flag bearer of freedom and progress to isolationist bully that wants to invade Greenland and become best friends with Russia?

American culture has lost its near-monopoly on optimism. We're now almost as cynical as the Europeans. (:D)

That cynicism means civic disengagement, technological doomerism and general symptoms of depression. That collectively degrades the mostly bottom-up structures we've long relied on, requiring shifts to less-efficient (and hastily cobbled together) top-down command structures.


>That cynicism means civic disengagement

By what metric? You have to go through Scandinavia, Germany, Spain, and a few Eastern european countries before you get to the US on voter turnout. Not to mention labor unions striking as a form of political protest (eg Italian labor unions striking against the Gaza war). And depression prevalence also seems to be higher in the US. Did you mean worse than Europe instead of "almost as" bad?


Fair. Maybe it's fatalism I'm mistakenly proxying for cynicism?

I disagree that the loss of optimism is causative. I've lost my optimism because I've seen what we're capable of as a country and it's not reassuring.

It hasn't been a short time, it's been a gradual process. I would look for example at the trust in congress over time.

https://www.pew.org/en/trend/archive/fall-2024/americans-dee...


America has never been that. With everyone having access to apps like TikTok, the brainwashing stopped working as well and people can see that it isn't.

> with everyone having access to apps like TikTok, the brainwashing stopped

The ad-powered social media addiction stopped brainwashing?


TikTok was a Chinese app that had no qualms against showing Americans what their country does.

> TikTok was a Chinese app that had no qualms against shooting Americans what their country does

If one wanted deeply pessimistic takes on America and Americans, there has been a media market for that since at least the advent of cable news. Mistaking TikTok, one expression of a phenomenon, for the general trend is mistaking a tree for a forest.


There has historically been a lot of US-critical content manufactured by the US, which normally deflects criticism towards individual failings, external enemies, or surrogate political effigies. A few examples: Falling Down, Bulworth, American Beauty, all the other 90s/00s media critical of suburbia, 24, The Daily Show. Most of these are left-coded, I’m not as familiar with the right-coded stuff (I usually tuned it out) but from what I can tell it’s usually aimed against foreigners and weak/effiminate liberals. 2010s/20s race activism made “white people” into the effigy for the first time, but that’s still a deflection.

Pre-sale TikTok was the first time that a mirror was held up to US politics from a global perspective, where the masses could get a less fitered and channeled understanding of how they are seen by the world. (Reddit provided this previously but it has fewer users and less impact.)

As much as I hate TikTok and short videos, it had a big impact. There’s a reason that they forced the sale. Domestic control of mass media consumption is the primary method by which public opinion is shaped within the US.


> which normally deflects criticism towards individual failings, external enemies, or surrogate political effigies

And next to that is the ever-profitable imperialist/capitalist/inherently-racist pigs content.

> the first time that a mirror was held up to US politics from a global perspective, where the masses could get a less fitered and channeled understanding of how they are seen by the world

This was happening simultaneously on other tech platforms. Moderation varied. But I think a lot of people are mixing up the tail and the dog in terms of which way causation flowed.


Cable News America-criticism? Is that when a gaggle of millionaires complain about either the racists in Middle America/how liberals smell like Europeans? Then someone blurts out, Dangit, We’re Americans, We’ve been slipping these pasts five minutes at being the bravest and most freedom-loving people, God Bless. To roaring or prompted applause.

Right, but then TT control was transferred to the US, and overwriting any lingering memories with Approved Propaganda is trivial and undoubtedly happening already.

America has absolutely been all of that and more within living memory. The problem is it's getting to the point where you need to be pushing 50 to remember a time when this was the case.

I think I would love to test it, but on the Pro plan I just did two small sessions with 4.6 Sonnet and it consumed my 5h quota within one hour.

The tokenizer changes seem to indicate that 4.7 isn't just a checkpoint but rather a model trained mostly from scratch, right?

You can change tokenizers without a complete retraining from scratch.

> — billing alerts fire in hours, damage happens in minutes.

And why do you need to use AI to tell us that. How much shorter could the prompt have been?


You are getting down voted but the first thing I thought when I read the above comment you replied to was that it was written by an LLM as well. It has all the stylings of it. Word choice, sentence structure, phrasing, metaphors, etc.

I think after the TTL expires the session should be autocompacted and the user should given a choice to continue with compacted version or be hit with the full read cost of continuing with their large but expired context. At the moment users are blind what is going on.


Consider applying for YC's Summer 2026 batch! Applications are open till May 4

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

Search: