Literally every single HN thread these days mentioning big companies the top comment is some cynical conspiratorial nonsense. Say Big Tech is out to get you in any way possible no matter how outlandish and people eat it up. Nothing can just happen for good or neutral reasons it's always tied to some master plan.
This is probably why they are launching now. There's no real other competitor right now in the nicely made managed openclaw space for personal use (or at least, very very early. Maybe Grokbot can be considered the lead competitor but afaik it did not gain mindshare). Instinct is supposed to come out soon and by not being associated with Meta it'd be really appealing just on that front vs this. But being first to market is a real reason to use it and chance to gain lasting market share.
The best selling point is an end-to-end purchase flow that you don't have think about at all, it does the price comparisons for you. However, if you want to use it to just get the list of top candidates and compare and go make the final decision yourself I'm sure it can do that for you too. Either way I see a future where these claw-like agents can fit into everyday use cases.
HN isn't really a good place to have an informed discussion on ads efficacy unfortunately. In this thread we have many many people talking about how online ads in general are completely worthless. All of the top comments I see are harping on things like Google Ads having bad targeting for reasons like you are getting recommendations on products you just bought (having experience in the ads industry, yes this is actually a good signal that you might buy another one, even if you don't want to believe it).
Meanwhile, the article is actually discussing ChatGPT ads specifically compared to Google Ads, which they do have success with.
Anyway, my guess is that it's a combination of 1. OpenAI being new and not having the massive work needed yet to set up a fully successful ads network, but they're recruiting a lot of talent from the area + recsys so it's probably not too far off. And 2. we don't really know in general yet how best to do in context ad recommendations. I think this will also improve pretty rapidly with how much talent is working in the space. The best ads will probably look a lot more like Google Search per-request targeted ads rather than Instagram style per-user targeted ads, but I believe we'll see lots of gains fast.
Nah, I think this is a common misunderstanding of how LLMs work, where people think that they mimic the pre-training data. Stylistically everything you see is an artifact of post-training, which is from reinforcement learning not from absorbing mass amounts of text. At some point a person or more recently a bot gave a thumbs up to an A/B tested response including em-dashes and claudisms galore.
Nothing is priced according to what they really cost. Things are priced at what the market will bear based on supply and demand. You might as well ask this question for oil, iphones, bitcoin, restaurants, or housing the answer is all the same.
That is the point though, if labs are maximizing svg image generation capabilities it is a very good thing. That's a general skill that is useful. So assuming they aren't specifically maximizing pelican bicycle svgs (and it doesn't look like they are) then incentives are aligned that the "benchmark" is measuring a general desirable capability.
How much infra have you seen of the top 100 sites in the world? I've worked on multiple top 10, and absorbing a 2x increase (and the peak is very likely more than 2x) is a very very hard problem that would cause hundreds of pagers to go off and load shedding to very high degrees. There is just not tons of unused capacity lying around in wait at the scale of github. "No one takes chances with critical components" is also very wrong for the simple fact that you don't know which is the weakest link in the chain until it fails.
RSUs are great for workers because if the value dips 75% you can just leave and get a new job and get new RSUs at a reasonable price. Of course it's not trivial to switch jobs but at a 75% cut it's surely worth at least looking around. I don't think you can really consider it a fault of the economy if you didn't.
Note: if your reason for switching is purely uptime, you will be sorely disappointed with codeberg
Their own site reports the 2 week uptime at 1 nine https://status.codeberg.org/status/codeberg and I suspect if you use the (really terrible) across-all-product-offerings uptime methodology that people love to post for github it would be a 0 nines overall service.
The millisecond-term zero sum game is part of what allows for a positive sum long term. For example, zero fee trading was pioneered by Robinhood and only possible because of payment for order flow, and as a result it's virtually unheard of now for retail to be paying per transaction. Now more retail investors can participate and everyone benefits. You can also point to lower spreads and faster execution as direct benefits.
To be clear, since the mobile app users are decreasing, what you would actually need to claim is that "web users are more likely to continue using the app than mobile users are", not just that "there are a lot of web users", since the base assumption is that web and mobile users use the site in similar ways. I don't really see any reason to believe the former, and actually having worked on social media extensively I think there's a strong chance that the opposite is true, that if mobile users are declining web users are declining even faster. There's a reason sites push mobile apps so much and it's because retention is normally far better.
Reading the responses to your comment the discussion would be a lot more productive if you shared your logs (preferably several of different top models since that's what you're claiming) where LLMs fail at this. Not very useful for people to go back and forth speculating on what you could have asked and with what formulation. As it stands for me simdezimon's logs are pretty definitive that there shouldn't be any problem for current capabilities agents to solve the task.
Databricks is most certainly getting charged API pricing no matter what harness they are using. OpenAI and Anthropic models are so sought after right now that they set the terms even at the world's biggest companies, there is not a chance to get a special agreement for subscription pricing.
You have no clue what you're talking about while trying to act smug and talking down to someone who knows at least a little bit more than you. Netflix cannot run without massive DCs, whether that be theirs or Amazon's.
So yes you would in fact need to pay ten times that and then some. And by virtue of being on hacker news I'm guessing your actual ads targeting value is more like 20x that or more.
2 main problems with this article: one it's fundamentally asking for something impossible, and two for the part that _is_ possible it's using the wrong tool for the job.
It's not possible to magic details into existence that weren't there. So if you think a color was wrong, well, that is not something that can be solved other than by using a time machine or finding a separate reference picture. Any human restoration artist would be doing the exact same thing of guessing the color.
The second problem brought up is that ChatGPT is willing non-color details into existence that aren't in the original picture (eyebrows etc). For this it's simply that the public ChatGPT image generator wasn't made for this task. You'd have to build a custom harness but I'm sure it can be done.
It's not hidden at all. Financial blogs very accessible to laymen like Matt Levine's Money Stuff have talked about this structure months ago. If you are an investor and surprised by this news you weren't sufficiently prepared and shouldn't have been investing in the first place.
I would guess the price for payment processing drops tremendously when you don't need to handle fraud or chargebacks at all. It's for sure worse for consumers but I do think it has its place (for example adult content that has huge chargeback risk or straight up blocked by other processors).
Your experienced latency will be heavily influenced by which DC you hit, in which case we care about the relative traffic loads because you aren't just hitting your closest one all the time.
Most of big tech's major data centers are in Loudoun County, VA on the east coast not the west coast. It's centrally located to be great latency for the east coast and OK latency for west coast and Europe. Plus a friendly regulatory environment and lots of existing DCs (AWS us-east-1, Azure East US 1/2)
If you're feeling any better latency on the west coast it's more likely to be placebo than real.
If the agent you are running really wanted to it could easily find a way to mount the windows folders and read them all. WSL isn't a security boundary, you are only barely more protected than people running grok in their home directory.
Wealth is not in fact finite. If you take a piece of lumber and carve it into a dresser you've just generated wealth. Resources kind of are but not really, extraction takes it from unusable to usable and essentially creates it (also, trees literally grow out of the ground). Similarly Tesla has made a lot of cars that people drive and AI companies have created products that have hundreds of millions of DAU.
Whether that "justifies" the market prices is orthogonal, all the market price says is that people want to own a piece of the pie at the current price being offered. It can stay that high as long as there continue to be people interested in owning shares of the companies.
You are poisoning the context and priming it to say load-bearing. I don't really think there is a good way to actually ban specific phrases other than have it do a second pass, it's something baked into the model from post training.
Within the 1982-2026 span there is an equally negative 3.5 sigmas deviation somewhere (no year labels on the graph). The article doesn't touch on it at all so I have no context as to what it could be. But it definitely suggests 3.5 sigmas is not really 1 in 7000 years.