Hacker News Reader

Comment by mtmail | Hacker News Reader

Can we speculate if X's CTO didn't like the reorg? 3 days ago "X’s director of engineering, Haofei Wang, has left the company " https://news.ycombinator.com/item?id=43470613

View on HN

More likely Musk liked xAI's CTO more than X's CTO, and parted ways with one of them. Which is his prerogative, as the owner of both companies. I'm sure Haofei would love to be the CTO of one of the strongest AI models in existence. I would, if I were in his shoes.

"strongest AI models in existence"

[citation needed]

To be fair they said one of the strongest models.

We're at the point now where any heavily monetized player can build one of the strongest models. I mean, especially with all of the incredibly open Chinese companies releasing cutting edge research now that most American firms have gone quiet or only releasing their tailings.

Technically my toy Markov chain I trained on Gutenberg books is one of the strongest models.

If you used real chain it definitely has the most strength.

See the leaderboard tab here: https://lmarena.ai. Or just try it. It's excellent. Usage is free of charge so far. Subjectively, on my prompts, it seems substantially stronger than GPT 4o and 4.5. I also like its more "informal" tone, and its willingness to discuss the more controversial topics.

I think it’s important to not support products by people who want to destroy democracy and who openly sieg heiling or claiming Hitler wasn’t responsible for the Holocaust.

On the bright side, Google's new Gemini-2.5-pro comes with no Holocaust-denying issues, and is definitely a lot better than Grok, and is also free and fast.

(I've been using it the past 2 days and am pretty impressed - the first high-end Google model I'd willingly use instead of GPT-4.5 or Claude-3.7.)

Meh. Google can keep its censorship and algorithmic narrative steering for itself, IMO. Quis custodiet ipsos custodes, or something like that.

The Gemini models have become notorious for being rather uncensored and unrestricted, with an oddly base-model-like feel, since at least last November.

Good to know, I will give them another try, thanks. Cooler heads might have prevailed over there.

Agreed, but that's not really related to the question at hand.

In normal times I wouldn’t have done this.

But we are at a point in history where everyone who believes in upholding democratic values needs to call out those in power who are actively trying to destroy it.

Vote with your wallet.

And it’s risky to build your product on top of Grok’s platform, because its owner is totally unstable and you don’t want to be exposed to ketamine fuelled decision making that could negatively impact your business.

If you use the free version, you're basically draining his money.

Ask any mainstream U.S. model anything remotely sexual and they all abort. Musk et al. claim they want free speech and no censorship, yet the U.S. evangelicals have their teeth so deep into the tech, that they block it or debank anyone who even remotely suggests women should have a pleasurable sexual experience.

Yep, it’s pretty smart

https://www.theverge.com/news/617799/elon-musk-grok-ai-donal...

Reply on news.ycombinator.com