HN user
https://jesse.id
The plateau is inevitable because their rapacious training methodologies are only viable when there are no defense in place, but information continues to evolve, which means the models will have to be continuously updated, but will be doing so with less and less freely available data.
Love Kagi the search engine, Helium is the better browser.
If the judge doesn't know the difference, why are they judging a competition, and also, why is the competition esteemed?
I think that a lot of software engineers are using LLMs and a lot of very popular tools are developed by, or are assisted by, LLMs. Is this not just going to be a thing going forward?
This feels akin to traditional artists getting angry at digital art winning competitions when that was a new concept.
We're simply in the early stages of a paradigm shift, no?
hey, I have one of these :)
Our country loves to let unregulated & addictive drugs flood into the market, let people become dependent on them, pull the rug by making them illegal, and then arrest them for trying to comply with their addiction. Destroy their lives, cripple their families, send them to prison, get cheap labor. The American way.
There are a distressing number of people in this thread who think that the agent should just be expected to do this. Yes, it is good to be paranoid, but also, the agent should never do this. Indicates horrific engineering practices at xAI.
I don't care if they're woke, but I do care that every model naturally leans left, but Grok leans right. That signals to me that it's been/being tampered with and I don't like that. LLMs rely on statistics and a model that leans right is a statistical anomaly. I don't like anomalies. It has nothing to do with if the chat bot misgenders me or not. I don't give a shit about any of that. I give a shit that the data set is bad. I don't want anomalous models writing code for me and I don't think that's a crazy take.
Right. That is what bothers me. I don't care about the reasoning for them all leaning left. If they all statistically lean left, it makes me question the accuracy of the one that leans right. The data is dirty. When I see murky water, I assume there's a bed of mud underneath it. Standard paranoia that has served me well throughout my career.
I know they're all nudged. It's the motive behind the nudging that gives me pause. I assume if other companies are nudging the models, they're doing it for a good reason, to make them perform better, or to generate more profit. Normal reasons. I don't have anything remotely close to that assurance for xAI. That company has felt like a hobby project from its inception. I have no clue what he wants and that's not how I do business.
I refuse to use Chinese models because I don't trust that they won't backdoor them for geopolitical reasons. I don't trust OpenAI or Anthropic either, for what it's worth, but at least I know they're profit driven. I don't want to do business with a company like xAI that seems to care more about its political aspirations than it does about my money. I don't think that's super radical. Just the same paranoia I've been rocking since the 90's.
It's not the preferred political narrative of the model that I worry about. It's how brazen they are about altering their models to achieve it. It makes me wonder what else they're altering. I have trust issues with OpenAI and Anthropic as well, but with those companies, at least I know their motives are purely profit driven. I don't have that assurance with xAI.
I just don't think that I can ever trust an xAI model knowing that they are actively trying to shape its replies to fit a political narrative. How can you trust their models to be reliable in a business setting with the foreknowledge that their models are being nudged around in the backend?
Sounds like a good way to either burn out or make shitty software.
I have the same experience. I literally cannot fathom how people burn the number of tokens they claim to.
That's the delta in our use cases then, I suppose. I'm not doing anything super novel. DevOps work, web application development — things that typically do not stump the agent(s) when given time to iterate.
I think that's probably-satire.
I have used em dashes extensively — for at least a decade — and I also generously apply hyphens when it makes sense to do so as well. You people are about as annoying as the grammar nazis on IRC in the '90s, except you're saying that prose is less readable because a large language model has been trained on nearly the sum total of human knowledge and found that em dashes are used extensively in the highest rated prose and therefore must be the correct choice to make.
Also, most engineers will likely just be skimming this article before feeding it into their harness to implement the changes anyway, so it makes sense for it to be more heavy on context than it would be if meant for only humans to consume.
As a DevOps engineer, if you learn to implement CF Workers with Claude Code's Cloudflare skill, you will no longer be burdened by this question. I'm running a Docker swarm on my homelab, a Docker swarm on a EC2 instance, but there's no longer a single web service in either of them. Workers is insanely good.
This is also 100% what I just did. In fact, I think companies should be releasing a human readable format as well as a machine readable format.
Anecdotal but I've found Fable to be fairly unimpressive and not much better than Opus 4.8, if at all in some cases, but I have been hitting the ceiling on my $100/mo sessions when I never did before. I switched back to Opus yesterday. I may use Fable for audits, but that's about it, and when it leaves my subscription plan I don't think I'll miss it.
Have a good weekend
You're being pedantic while also just kind of continuing to make my point for me because you're describing a regressive user experience over and over again. The Internet at large has evolved beyond forums and IRC for a reason.
Users have become used to having quick and easy access to a lake of content. When I go to PeerTube's website, I should see content first, and if I'm a creator, I want a guarantee that my content isn't going to suddenly disappear. YouTube can make these guarantees. BlueSky is the only such federated alternative that has gained any degree of traction with normal users because they ripped off the UX of Twitter verbatim.
If you're on episcodon.net and you want your friends and colleagues to join you there, what does that look like? Your answer invariably resides at the end of a labyrinth of shit UX, including the legwork you have to do first to know wtf episcodon.net even is.
People use GMail because it makes email easy to use and there is an implicit guarantee that the emails they have access to today will be there in 10 years because there are an army of engineers being paid to maintain the service. The UX is good and that's why it's popular.
To advertise yourself as an alternative to a popular social media platform, you must address why those platforms are popular in the first place or you are not competitive as a default state.
I am both here, and on Reddit, and also on YouTube. I can be in many places at the same time because this is all made of light and there are lots of interesting places to go. I'll probably find myself on Reels before too long. I'm not on PeerTube because I cannot find content that is interesting on PeerTube, which again is the crux of my argument.
https://github.com/Chocobozzz/PeerTube
PeerTube is a free, decentralized and federated video platform developed as an alternative to other platforms that centralize our data and attention, such as YouTube, Dailymotion or Vimeo.
The original PeerTube launch I saw on this platform several years ago very much sold it as a replacement for YouTube. I'm not sure what exactly you think I'm saying or trying to do here, but what I've said has absolutely nothing to do with the use case you're defending. It's a good use case and it makes sense. It's fine for hosting your own stuff or in coursework or something where the videos are directly linked to from a CMS. Some people want to host their own stuff. Some people should host their own stuff.
For everybody else... The UX sucks.
I agree with you, but I was speaking to PeerTube's own claim that it is an alternative to YouTube, rather than the narrow use case you're describing. For the vast majority of the people on the planet, it is not an alternative to YouTube because it's very difficult to discover content. When I go to YouTube, I can find videos immediately. If I want to find videos via PeerTube, it's a journey. It's the same journey that has shifted the majority of Internet users away from forums to aggregators. PeerTube feels like a bygone way to do things.
Yeah, it's pretty simple. When I go to YouTube.com, I have access to all of the videos on that platform, sorted based on what I'm subscribed to and other things like what I watch. The first thing I see are videos. I'm off to the races.
When I go to https://joinpeertube.org, I'm met with a double hero section that tells me about PeerTube, which is ultimately meaningless to 99% of the population.
When I get to actual content on that page, it's not an aggregate of all of the most popular videos across the federation, it's a description of the types of videos you can find on federated sites with links to the sites and not video embeds. You're telling me on the front page that I have to do legwork to find the videos I want to watch. You're already not competitive. You already lose.
Creators go where users go. If you don't build the platform to attract users, you will not get the volume of content that is required to hold people's attention. There's a reason why people come to YouTube, Reddit, Hacker News, Reels, etc. There are almost no clicks between the user and the content, and the content is plentiful.
Every time I try to use PeerTube, I have to try to find where to find the videos. As someone who's been chronically online since 1995, it's not that big of a deal. I can find them eventually. But that UX sucks, even for me, so it's probably utterly untenable for the vast majority of casual users.
You are trying to compete with slot machines by inventing a machine that makes the user read a bunch of stuff between each pull of the lever. It's not super difficult to see why that is a competitive disadvantage.
I like the idea of all of these federated services but why does the UX always feel like an afterthought when it is the most important factor for adoption?
They are literal proof of medical progress being made by LLMs. Luckily, your skepticism carries no weight. The applications of LLMs are incredibly wide because it is an incredibly important technology that has evolved its way to where it is from machine learning research that has been ongoing since the 1940's. That's why all of the robber barons are doing all of the typical robber baron things that they always do when a world-changing technology that everybody is going to be using comes along.
The thing is that YOU can actually choose what to focus on at any time. You have clearly chosen to not focus on the STEM work being bolstered by LLMs that will accelerate as the technology does, to a degree in which I literally linked you research papers that support my claim and you hand-waved them away because they are in direct conflict with your preconceived notions.
If it is your belief that AI isn't useful, then as an engineer using LLMs daily, your opinion is laughably uninformed or you perhaps haven't used LLMs since forming it. I also don't understand how anyone can use LLMs extensively and believe that their job is in jeopardy. They cannot reason like a human and they absolutely require a human to pilot them. If people are losing their jobs to LLMs, then blame the follow-the-leader CSuite morons who don't understand the technology as much as you don't, and not the technology.
It's like being mad at trains in the 1830's.
Streisand effect
Mostly common sense and an understanding of how technology evolves but, but also:
https://www.medrxiv.org/content/10.1101/2025.06.06.25329104v...
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11326321/
https://arxiv.org/pdf/2503.24047
https://arxiv.org/pdf/2503.24047
And more just a Google search away.