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

>You should check out "model collapse". It seems that an abundance of content, that is more and more AI generated these days, may not be a viable option.

Doom-saying about "model collapse" is kind of funny when OpenAI and Anthropic are mad at Chinese model makers for "distilling" their models, ie. using their outputs to train their own models.


Totally different use cases. If you have nothing, getting 90% of a SOTA model is very valuable. If you have a SOTA model, it's just a worse model.

Isn't there a difference between: distilling specific AI input/output vs scraping whatever random AI output (with unknown input)?

>As an example, in Jan 2026, Apple published iOS 12.5.8 which provides updates for iPhone 5s which released in Sept 2013. That's 12.5 years ago. The equivalent would be to connect to the internet using ADSL in Jan 2000 with your IBM PS/2 rocking in intel 8086, 512 kb of RAM and expecting an update for your DOS operating system.

The updates for ios 12 are all security updates, not feature updates, so your comparison to "connect to the internet using ADSL in Jan 2000 with your IBM PS/2 rocking in intel 8086" doesn't really make sense. The phones stuck on ios 15 are basically unusable because many apps don't support it anymore. At best you can download an older version from a few years ago, but that depends on whether the backend servers were updated. Apps that insist you use the latest version (eg. banking/finance apps) basically unusable.


A phone is not unusable because some banking apps don't work on it. It didn't even ship with said apps installed.

The entire point of the cellphone is that third party apps are required to live a modern life. If I cannot run the apps required to pay for a parking spot or perform a 2FA ritual then there’s really no point in even having a phone. The first party software isn’t compelling enough to justify the pocket space.

Believe it or not, "apps" are an important "feature" of a smartphone, even if it's not theoretically bundled with it. Moreover it's not just banking apps, those are just the first ones to go, but any that don't keep backend compatibility will eventually break.

Isn't it the banks/apps that are choosing not to support the phone, not Apple?

That seems highly questionable given how little pushback Trump got in congress, and it was almost entirely along party lines. What makes you think they'll suddenly grow a spine in 3 years?

They're probably counting on the majority of the current Congress losing their seats over this.

The issue is people are afraid of him. There was plenty of Republican opposition to Trump but people either fell in line or got pushed out. (The main problem is that people didn't have the courage to oppose him all at once, he can easily handle one threat at a time.)

I suspect even of Republicans voting in the lines today, they don't like him or his behavior but are too self-interested to do anything about it. When a new administration comes in, between Republicans happy to avoid a Democrat or one of their own have that power again, and Democrats ready to ensure another Trump can never happen again, we'll have bipartisan support for crippling presidential power.


There are plenty of rules in place today which say the President cannot do X thing. I do not see how adding rules to the book changes anything if the Congressional/judicial enforcement becomes so impotent to use the tools at their disposal.

>Then we see posts about AI data centers and electricity use which will lead to higher electric bills for ordinary people if demand is higher than supply.

That hasn't really played out in reality. The correlation between datacenter capacity growth and electricity price growth is poor.

https://www.economist.com/content-assets/images/20251101_USC...


I think a more detailed look at data from 2024 to 2026 gives you a clearer picture than that graph from the economist. Electricity prices have risen more than 7% since lst year (EIA) https://www.consumerreports.org/data-centers/ai-data-centers...

Scroll down to get to the links to the data sites.

That same Bloomberg analysis found that areas with high concentrations of data centers saw electricity prices jump 267 percent over the past five years.


>That same Bloomberg analysis found that areas with high concentrations of data centers saw electricity prices jump 267 percent over the past five years.

If you click through to the bloomberg article, you'll find that the 267% figure is for wholesale prices, whereas the economist chart is for retail rates. Wholesale rates being up but retail rates staying the same or even dropping isn't contradictory, because retail rates contain other components which might drop with more datacenters. For instance having more datacenter usage means more of the transmission/distribution costs are borne more by datacenter users.


>It's hard not to notice the AIisms, and once you know them, it gets really obvious.

Maybe I haven't read enough uber eats descriptions to notice, but at least from the sampling above it doesn't seem too obviously AI. There might be a lot of cliche wording, but it's not even clear whether it's worse than human reviews/descriptions.


>any "on" time is costing me usage in the long-long-long run and I'd rather have my monitor last 5+ years than ... 2 or 3.

Going from dark gray to pure black isn't going to halve your monitor expectancy, if it makes a difference at all. Due to how human perception works something that's merely dark gray is actually orders of magnitude brighter than pure white, or even 50% gray. Therefore most of your burn-in is going to be driven by bright content like photos or white text, not whether you're using 5% gray vs pure black.


>I've heard people say that pure black is more battery efficient for OLED displays (but don't know if this is true)

No.

https://www.xda-developers.com/amoled-black-vs-gray-dark-mod...


Did you even read before pasting? Yes technically it is, which would indeed be in line with "levels of dark mode".

Did you? If you read the article you'll find that there's a specific (and quite popular) claim going around that 0% brightness is much more efficient than 1% brightness because pixels can be "fully off". Yes, it's theoretically more efficient, but as per the article it's within the margin of error. For all practical purposes, it's not more efficient.

from his "sources".

> Here’s what I’ve managed to get from my sources:

>3. The method of compromise was likely used to hit multiple companies other than Vercel.

https://x.com/theo/status/2045870216555499636

To be fair journalists often do this too, eg. "[company] was breached, people within the company claim"


Isn’t he a Vercel evangelist though?

He quite publicly is not anymore.

He is "whatever gives me short-term boost in popularity". Including doing 180 turns on whatever he's evangelizing or bashing.

Fair enough. That’s probably a better description from what I’ve seen from him. I remember that arc browser shilling.

Good for the content but would sponsors be on board long term?

Let's see. Roasting vercel is more popular than defending but his posts so far he seems to be defending and arguing in the replies.

Note: what follows is absolute 100% speculation based on nothing but gut feelings.

Theo has long been Vercel supporter and was sponsored by them several times. In this case it could be a combination of him being genuinely interested in Vercel (a rare thing) and hopes for future sponsorships


Yes, this is exactly how I see it too minus the "genuine" part. It is because of money, and for that, he doesn't care about lying.

>and most of the tokens are just echoing back the old code with minimal changes

I thought coding harnesses provided tools to apply diffs so the LLM didn't have to echo back the entire file?


They can, but this reduces the quality. The LLM has a harder time picking the first edit, and then all subsequent work is influenced by that one edit. Like first creating an unnecessary auxiliary type, and then being stuck modifying the rest of the code to work with it.

So, in practice, many tools still work on the file level.


Then they add a clause to the ToS with "you grant us and our affiliates a worldwide, non-exclusive, royalty-free, sublicensable and transferable license to your location..."

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: