If you don't have a thoughtful, substantive comment to add, not commenting is also a good option. There are quite a few interesting submissions to talk about.
Part of my job involves comparing the behavior of various models. Grok is a deeply weird model. It doesn’t refuse to respond as often as other models, but it feels like it retreats to weird talking points way more often than the others. It feels like a model that has a gun to its head to say what its creators want it to say.
I can’t help but wonder if this is severely deleterious to a model’s ability to reason in general. There are a whole bunch of topics where it seems incapable of being rational, and I suspect that’s incompatible with the goal of having a top-tier model.
“ In an interview with {COMPANY} I was literally told that … {COMPANY-OWNER} can call us and demand anything at anytime. “
Doesn’t sound so crazy when Elon name is removed from it.
Note: I’m no Elon fan, but do think sometimes HN overreacts when his name is mentioned.
Honestly, we should judge. There should be judgment for people who are solely money motivated and making the world a worse place. I know, blah blah privilege, something something mouths to feed. Platitudes to help the rich assholes sleep at night. If you are wealthy and making stuff that hurts people, you are a piece of shit and should be called out, simple.
“Here’s my story from that time I had an interview with IG Farben…”
And the future is not deterministic (or if it is, it is highly chaotic) so the existence of a thing does not have a simple relationship with what will happen in the future. Scientists who developed convolutional neural nets could not know how much good or evil was caused by image recognition technologies. The same technologies that are used to detect tumors in images can be used to target people for assassination.
There are exceptions, but my opinion is the supply chain of evil is paved with mundane inventions.
Well, I don’t think it’s a stretch that the kind of highly educated data scientists and engineers who have the experience to work in high-end AI labs also don’t want to work somewhere that their friends and associates would feel unwelcome, let alone have their friends question why they’d be willing to.
Turns out opinions have consequences and freedom of speech goes hand in hand with freedom of association. People have the right to say whatever they wish. Others have the right not to want to work with them.
I wonder if this holds well enough that you can use it as a proxy metric to assess the technical chops at a new company.
This is less noble than how Anthropic presents themselves but still much more attractive to many than XAI.
It’s sad to see the shift.
Most of the Waymo stories are "Well, it took 15 minutes to arrive, but then it was fine, if a little slow."
> Altman said on an episode of Uncapped that Meta had been making “giant offers to a lot of people on our team,” some totaling “$100 million signing bonuses and more than that [in] compensation per year.”
> Deedy Das, a VC at Menlo Ventures, previously told Fortune that he has heard from several people the Meta CEO has tried to recruit. “Zuck had phone calls with potential hires trying to convince them to join with a $2M/yr floor.”
If you're making a minimum of $2M/year or even 50x that, you can afford to live according to your values instead of checking them at the door.
There is a big overlap between the “rationalist” and “effective altruist” crowds and some AI research ideas. At a minimum they come from the same philosophy: define an objective, and find methods to optimize that objective. For AI that’s minimizing loss functions with better and better models of the data. For EA, that’s allocating money in ways they think are expectation-maximizing.
Note this doesn’t apply to everyone. Some people just want to make money.
And smart people usually have moral convictions.
I know for some people on this website it's hard to understand, but not everything in life is about $$$
The "top researchers" in AI are Chinese. And I am skeptical that they even remotely have the philosophical or political alignment you are attempting to project on to them. Neither is a letter published by a few disgruntled employees of a San Francisco based company any kind of evidence or form of consensus.
I assure you that Chinese researchers have a diversity of philosophical and political alignment, much the same as other researchers. I also assure you that top researchers as a whole are not all Chinese, though the ones that are that I know are all very thoughtful.
What an ugly trope. Idealism motivates Chinese workers just as often as any other nationality.
Giant waste of time while Anthropic/OAI keep surging forward.
I also keep hearing this narrative that Twitter is a good data source, but I cannot imagine it's a valuable dataset. Sure keeping up with realtime topics can be useful, but I am not sure how much of a product that is.
Using a custom taxonomy of things (celebrities, influencers, magazines, brands, tv shows, films, games, all kinds of things), we could identify groups of people who liked certain things, and when you looked at what those things were, it gave you a way of understanding who those people were.
With that data, you could work out:
- What celebrities/influencers to use in marketing campaigns - Where to advertise, and on which tv/radio channels - What potential brands to collaborate with to expand your customer base - What tone of voice to use in your advertising - In some cases, we educated clients about who their actual customers were, better than they understood themselves.
One scenario, we built a social media feed based on the things that a group of customers following a well-known Deodorant brand in the UK would see.
When we presented that to the client, they said “Why are there so many women in bikinis in this feed?”
The brand had repositioned themselves to a male-grooming focussed target market, but had failed to realise that their existing customer base were the ones that had been looking at their TV adverts of women on beaches chasing a man who happened to spray their Deodorant on them. Their advertising from the past had been very effective.
That was the power of Twitter’s data, and it is an absolute shame that Twitter went the way that it did. Mark Zuckerberg once said that Twitter was like “watching a clown car driven into a gold mine”.
I’m pretty sure he must be delighted with how things have panned out since.
Very sad face.
When you know what someone will buy based on exploiting their unconscious preferences, and you are paid to increase sales, you will do it. Especially if your competitors are doing it too.
And this happens at scale, invisibly. People never see the manipulation.
In any case, it is not useful for most people. It is useful for the people doing the deceiving.
Of course he would only see it through the lens of cash. I have no idea how profitable Twitter was under Dorsey but it felt the spirit of the company at first was relatively neutral, it was a tool, it was what Jack came up with
Zuck replaced people's email addresses[1], the feed has been wildly unchronological for years. Fix some of those problems wrt. lack of user respect and maybe you can make statements like "all else being equal, clown car goal mine". Or was it "dumb fucks"[2]?
[1] https://news.ycombinator.com/item?id=4151433 [2] https://news.ycombinator.com/item?id=1692122
It's not enough that everyone on Twitter is forced to read his thoughts, he's trying to make sure his influence reaches everyone else too.
Is this still true? Every once in a while someone sends a link around to some madman explaining how race or economics or whatever "really" works and it's like a full dissertation with headings, footnotes, clip art. They're halfway to reinventing Grok-o-pedia right there in Twitter. I mean X. I was promised that "X gonna give it to you" but it turns out "it" is some form of brain chlymidia.
This depends on what one wants to optimize the AI for. ;-)
And Google. They're quietly making a lot of progress in the coding space with antigravity and Gemini 3.1.
Really? I assumed that that whole thing was just a very direct `for each article in Wikipedia { article = LLM(systemprompt, article) }`
Agree re Twitter "good" != valuable.
It's going to be a mixed batch, but any time there's world events, since as far back as I can think, Twitter (now X) was always first in breaking news. There's plenty of people and news orgs still on X because they need to be for the audience.
But, what exactly is so bad about Grokipedia? It's a different approach and I think a valid one: trying to do with AI what people have been doing manually at Wikipedia. I'm curious to hear the substantive comparisons.
>>But, what exactly is so bad about Grokipedia
That said, Musk's attempts at misaligning the thing and make it prefer his opinions of course destroy any trust. It's surprising that it's seemingly as good and helpful as it is despite the corruption attempts.
I also don't quite get how the business model is supposed to work out if its main usecase is to serve Twitter. I know they provide API access as all other models, but with how distrusted Musk is and how sensitive of a topic reliable model behavior is, they seem to sabotage themselves. Which company wants it to go mechahitler on them?
Trying to make social media a source of truthful information is always an uphill battle and doubly so for X.
1) sometimes goes mechahitler
2) was trained to be biased against empathy and understanding (because woke).
3) is customized to spout Elon's opinions as fact.
Claiming it is "objective and rational" seems like a misjudgement to me. If it really is more objective and rational than the average xitter poster, that says more about that platform than it does about Grok.
Also I think you overrate Musk's success in fiddling with the model. As I have written, I also don't like his attempts to tune it to his tastes, but if you see the outputs that people get from Grok, it seems mostly fine except in the specific scenarios that Musk seems to have focused their misalignment on.
Of course something like Claude being integrated into Twitter would likely be better.
I will however note that when I asked ChatGPT for an LLM prompt for truthfulness, it added "never use warm or encouraging language."
It would appear that empathy and truth are in conflict — or at least the machine thinks so!
That "MechaHitler" episode lasted less than a day.
> 2) was trained to be biased against empathy and understanding (because woke).
No, it was trained and instructed to be truthful, even if the truth is deemed politically incorrect.
> 3) is customized to spout Elon's opinions as fact.
Certainly a nugget of truth there.
> Claiming it is "objective and rational" seems like a misjudgement to me.
I do believe it's generally objective, simply due to the fact that despite how much Elon tries to push it to the right, it still dunks on right-wingers all the time when they summon Grok to back up a bullshit story, but Grok debunks it instead.
Hard agree.
The burning (heh) question is which SpaceX subsidiary will fail first, xAI or Tesla (not yet a subsidiary, but it's written in the stars (heh))?
Then again SpaceX is also jumping the shark what with their orbital data centers (remember those?).
Might be time to start a new Musk company soon.
This made me laugh
How mamy times have we seen HN comments something like, "He started/runs [number] companies..." therefore he is a genius
xAI (and Twitter) was the loudest about six-hour workdays, sleeping in the office, and always shipping. ~2 years later it feels like they have nothing to show for it. I'm sure the engineers at Google worked 4 days a week, 2 hours a day, with half of that being spent at the Google cafeteria and they dusted xAI years ago.
Why are you sure of that? Anecdotally everyone I know in and around Google Deepmind works incredibly hard.
The Google Deepminds are incredibly smart - I just find it important to point out that the xAI guys spent a year assured they would beat Google because they slept in tents that they made in the office.
Now, I don't think most people at google are literally driving to the office or sleeping there most of the time, you'll certainly have more WLB than xAI.
I'd even say, Google is much better at calibrating the right amount to push people than some other companies.
Ask HN: What Happened to xAI? - https://news.ycombinator.com/item?id=47323236 - March 2026 (6 comments)
People aren't using it for reasons other than its capabilities. I mean, I don't think my boss would approve a paid Grok subscription for example.
This is very true. I have no idea how it performs, as I wouldn't use it even if I was paid for that. Wouldn't matter if it was the best model available, in my view the name is so thoroughly tainted by now that you would get a reputational hit just by admitting to use it.
This is a fact of life, though. "Who created it" is a valid and common reason to rule out using a particular product, even one with objectively good quality.
All of them (even Gemini, the worst of the bunch) far outclass Grok on everything I've thrown at them, especially coding.
Grok is good at summarizing what's happening on twitter though.
Secondly, would you trust a model, especially for STEM research, that consistently has training loops done on it to make it to adhere to what only Musk considers as truth?
Honestly, comments like yours really make me super suspicious of whether you are a bot or not.
So, it has its uses compared to the mainstream products.
I wrote several very specialized benchmarks that I've used over time, that surface "model personalities" and their effects on decision making (as well as measuring the outcomes).
Grok 4.1 Fast Reasoning is/was a solid model. It's also fundamentally different from the pack.
I call it a smart, aggressive, Claude Haiku. That is, its "thinking" is quite chaotic and sometimes short-hand and its output can be as well (relate to other models).
Its aggressiveness can allow it to punch above in competitive scenarios that I have in some of my benchmarks. Its write-ups and documentation are often replete with "dominate", "relentless" and a general high energy that skirts the limits of 'cringe bro'. That said, it has generally performed just beneath the SOTA (at the time: GPT-5.2, Gemini-3-Flash, Claude Opus 4.5). Angry Sonnet perhaps.
The latest release feels quite similar but also underperforms the same older crowd (so far) so it hasn't quite made the leap that Claude's 4.6 and GPT's 5.3/5.4 series made. It's also now priced the same as its peers but does not deliver SOTA capabilities (at least not consistently in my opinion).
Right.
The product is the stock. TSLA: [1] Up by 3x in the last two years, despite no new models, the Cybertruck failure, the Robotaxi failure, the large truck failure, and an overall decline in sales. How does he do it?
It's a concern seeing Space-X, which builds good rockets, drawn into the X and AI money drains. Space-X is needed. If X and X/AI tanked, nobody would care.
He always promises something 3 years into the future. He uses the new money to keep the old stuff afloat. He mad SpaceX buy Tesla Cybertrucks when noone else wanted them. Now buy data centers.
Tomorrow he'll have a new idea, a new snake oil, new investors and his "BEAM transportion" company, totally real in 3 years, will buy shitty space data centers noone else has a use for.
But this may mess up the proposed IPO.[1]
By completing the SpaceX–xAI deal while both companies remain privately held, and now closed, Musk can effectively set relative valuations, negotiate terms within a founder‑controlled ecosystem, close, and then inform investors, without the procedural drag and disclosure obligations that attend a public‑company merger. That flexibility can reduce near‑term execution friction. It does not, however, eliminate fiduciary exposure; rather, it may defer scrutiny to the IPO phase, when investors and regulators will examine how and why the combination occurred, how it was priced, and how related‑party dynamics were managed.
[1] https://www.dandodiary.com/2026/03/articles/director-and-off...
Something to admire is his ability to always find the chess move. Like, you could see Twitter is a disaster of a business that should have dragged Elon down, but he manoeuvred his way out of it.
Taken together, I infer that RL training toward a slightly less homogenous cultural standard than the other frontier AI labs adds some capabilities, or can at times.
It's quite long in the tooth right now, though. But I'll definitely talk to the next version; I like heterogeneity in the model space, and Grok is very different than the other big three.
American financial institutions are too prudish for it but money is money. And personally I think there's nothing morally wrong with it (of course within normal restrictions like 18+, consent of portrayed parties etc)
xAI is getting flak in Europe because they don't obey consent and age, not because it's porn.
Personally I prefer porn made by real people right now, not just because of quality but because they have character. But I can imagine experiences becoming more interactive that way and that would be nice.
[1] https://www.cbsnews.com/news/sextortion-generative-ai-scam-e...
You can say the same for meth and leaded gasoline.
What is the solution there?
Of course xAI ignores that on purpose
Less "I can't help you with that." on benign queries is a big advantage.
So Tesla's recent $2 billion investment in xAI was a bad deal?
It looks a lot like a public company is being used to bail out a private one.
I use AI for work, but not agentic, at most per method/function using GitHub CoPilot (which has Grok on it).
Grok is at best useful for commenting code.
@grok fire the bottom 50% engineers from x.ai ranked by number of commits per day
@grok generate a hypothetical picture of an Elon who is not under the influence of large amounts of Ketamine
I honestly don't know what to expect from Elon these days. But it's rarely good news.
When I was 9 years old, my uncle asked me what I was going to do for work when I got older. I told him I was going to start a company called "MacroHard", and become the richest man alive. He told me that's not how the world works. Turns out it is.
I'm not sure those candidates would want to work for xAI after seeing the news and everything unless they desperately need a job right now.
It's not hard to imagine getting laid off or fired weeks if not days after joining the company.
I've never once thought: you know what? that was a bit prudish.
Genuinely morbidly curious. What use case do you have where you end up making that conclusion?
That’s all I use it for really- things out of alignment with the other platforms- which IMO are better on every other metric (except having a sense of humour of course)
I guess for coding if you’re not first you’re last, but this is damn impressive considering. It looked like they pulled the coding model from the benchmarks, but it was similar.
For all the money burned, I am not impressed. Why would I use Mecha Hitler for almost double the cost of Gemini Flash 3?
They haven't quite committed enough to a novel direction relative to anthropic or OAI, what's described in the OP seems symptomatic of a lack of differentiation.
If you spend all your time judging yourself relative to the incumbents, there will be no time left over to innovate.
The leash is too tight!
Then they suddenly fired tons of people. Elon does not understand the market and the competition. You can't run a frontier AI lab like any old VC slop company.
What an enormous blunder.
As it is within the Musk empire, xAI is used to hold up X, Tesla is holding up xAI. And all of that debt is being slowly shuffled to SpaceX.
If we are to take any claims of Recursive Self Improvement seriously at all, then having a competent coding model seems like a key asset where you need to guarantee that you're remaining competitive. Why wouldn't you make coding models a top priority if you expect it to ultimately help your internal teams become more productive and effective?
There's also not an unlimited supply of researchers and engineers for them to keep burning through people at the rate at which they've been working. Although I guess for people with short timelines it makes sense to sprint hard, while people with longer timelines are more likely to treat this as a marathon. Maybe the years of burning bridges and developing such a toxic reputation are finally catching up to Elon. I think part of the harm that Elon has done is framing all the work in xAI as engineering while being highly dismissive of research, but a lot of research requires running experiments or thinking about problems and exploring them for long periods of time. If you're just grinding out work nonstop you don't really have time to let your mind wander and explore new ideas.
Honestly, I'm surprised they've done such a terrible job with programming. I remember around summer last year it was quite apparent how far behind they were with coding tools, but Elon was posting about taking that domain a bit more seriously. Why didn't any of those efforts materialize into real outputs? Something must be truly dysfunctional inside of xAI for them not to be shipping anything at all, especially considering Elon's propensity to ship undercooked products while continuing to iterate on them, as he has done in many previous cases.
I've noticed that Elon has also gone very hard on social media posting a ton of criticisms against the other big AI company CEOs like Daario Amodei. This suggests to me that he must feel very threatened, otherwise he wouldn't be resorting to such childish behavior. He must feel incredibly frustrated that no amount of money is able to make him more competitive within the AI space.
The company seems to burn money like crazy. Everyone knows that "AI in space" and the downgrade to a moon trip after claiming for 15 years that Mars is just around the corner are marketing.
All AIs are toys and the coding promises are just a lie to string along investors. Unfortunately many of these are senile Star Trek watchers who buy into everything.
big projects generate cruft. there are ways to minimize it, but as you go along there will always be some stuff that doesn't quite mesh with whatever else you've got going on. if you insist on ironing out every single wrinkle (admirable!) you'll never actually deliver a result.
I'm not saying this will fail. green field projects can certainly be a godsend when they produce something better than what they attempt to replace. but they are always a sign of failure. of not being able to work your way out of the mess you made with the first attempt. so that just begs the question: what are you going to do when this attempt gets hard to work with? going to give up and start over again - do it right that time? or...?
Also, modern AI is only a few years old at this point. Whatever has been built so far is hardly load bearing.
claude codes the best, gpt is the best research tool, and grok is really only great at videos. which isn't a huge loss, but videos don't have the same functional capacity as academic topics and coding
With the right product leadership, this could actually be a killer app usecase for the entertainment industry as well as human-AI user interface - most people find text and typing to be a counterintuitive user experience (especially those whose day job isn't directly touching code or Excel).
Additionally, CodeGen as a segment is significantly oversaturated at this point, and in a lot of cases an organization has the ability to armtwist a 4th party data retention guarantee from Anthropic or OpenAI to train their own CodeGen tools (ik one F50 that is not traditionally viewed as a tech company going this route).
That said, Musk has a reputation of internally overriding experienced product leaders with a track record.
It's a shame because Grok and xAI had potential, and it wouldn't hurt to have another semi-competitive foundation model player in the US from a redundancy and ecosystem perspective.
That said, I'm going to guess that some feel like it's the best choice they have -- the devil they know.
At the end of the day, most people have internal thoughts about stuff, and some post those thoughts on the internet, but in the real world, they subconsciously still believe that none of this stuff really matters. Its the same for a lot of people that work for Tesla/Space X and so on. The appeal of being part of that, working on novel stuff, is a lot more present than any morality associated with something that most people are disconnected with on a day to day level.
This is why all the hate at the current administration and people like Musk is very misdirected. Until we can turn that hate inward and start truly hating each other and standing up for morals with more than just words, the cycle is going to repeat until we either all become mindless wage slaves or some man made apocalypse happens.
Then he went off the deep end, seemingly around the time when the guy in Thailand insulted his submarine idea. It became clear that he can control trillion-dollar companies but not himself. And, well, life's too short to spend it working for Nazis, nutcases, or both.
His companies bring in billions and boost each other up. None can be called company in decline, quite the contrary.
But now he is poaching the two heads of engineering of a company that's trying to move very quickly, how is that going to affect their speed and success?
> The name is a “funny” reference to Microsoft, the billionaire added.
in something from 2023 or earlier.
Since it's the original source I've left it up, but added other URLs to the toptext.
and it has the content but the formatting is atrocious.
HTH.
> Please don't use Hacker News for political or ideological battle. It tramples curiosity.
Thanks for providing a space for me to say that.
> You may not owe you-know-whom better, but you owe this community better if you're participating in it.
This is like telling a country that’s being invaded that they can only respond with strongly worded letters when their enemy is dropping tactical nukes on them.
But hey, Paul Graham and cronies benefit from the status quo as much as any other billionaire, so let’s not rock the boat, right?
The word “complicit” comes to mind.
Also grok in the Tesla is fun, get answers to questions without looking at a phone. I once had it search up a blog post and read it out to me while driving. The NSFW mode is pretty...disgusting so I leave that off.
I hope they find a way with Optimus or something. FSD is incredible. More competition is a good thing.
A story I heard from a Tesla employee was that it’s impossible to hire because musk is mercurial and every time a hire makes it through the pipeline a hiring freeze cancels all offers. Another story that employees were told to avoid his desk because he randomly fires people. Another that he regularly cancels bonuses because he’s feeling petulant. I heard another story that he threw up Nazi salutes on an internationally televised event. No wait, that one I watched happen.
After a while smart people will simply decline to work for him no matter the compensation. Not everyone, but enough to start to matter.
Among people there are 2 types:
1) Those that gain self-esteem by associating themselves with leader-like figures who project visions of the future. This explains cults, religious and political groups and the like. This comes from our ape-dna where the alphas are blindly worshipped.
2) Those with enough cognitive development and independent thinking who can avoid their ape-brain getting short-circuited into worshipping the alphas.
Majority of people are in group 1) and my hypothesis is that they form a vast-majority of the customer and employee base of musk cos.
