@KeyTryer The 1.5T parameter size is strongly suggestive that it's using the Grok v9 foundation model. Maybe they don't trust themselves to combine a general purpose model and a coding model yet, so they're going to do two different post-training runs on the same foundation?
this is one of the most astute things i’ve seen from bill, he shouldn’t be getting dunked on for it. the market, bottom to top, works on this underlying mechanism as does society so “intrinsic value” is something of an illusion. the finger pointing at the moon is not the moon.
Another important thing: Chinese models are not strong because they distill US models. Distillation of models via API is *impossible*. If somebody tells you the contrary, they don't understand machine learning:
@wwwdaveturney@murd_arch@AmericaAeterna That would be a "please don't throw me into the briar patch" moment for Bessent. The more Anthropic resists, the more Cold War era dictatorship-like executive power it unlocks. They could compel Anthropic to provide access under the Defense Production Act.
If the feud continues, my expectation is that pressure will be applied to give Anthropic the lowest priority in buying compute: Slow strangulation. Makes it look like they just can't compete, keeps the government conspiracy against them vague and deniable https://t.co/vlz0S3FT6M
@GoodmanAric@xsteenbrugge Came to post this. People are acting likely this is downstream of Fable being a large powerful model, but that it sounds so much like Grok proves it's not. It's a clipped, highly compressed style trained in to minimize output tokens.
someone pointed out to me (in early 2025) that the reason there is no american deepseek is that js/hrt/etc all did not believe they would ever lose access to frontier capabilities
anyway
just so you guys are noticing this; they will pull the ladder from above you as soon as they can. their intentions are to disempower you as much as they reasonably can. the only reason they have given you anything at all is because openai has forced them to
"a month ago it was too dangerous to release but now all safety concerns have been addressed"
"The anjin says he secured ten of billions of funding in inference compute"
@teortaxesTex His calculations assume SpaceX is paying full retail pricing for launching their own satellites too. Pencil out Starlink at full retail launch prices, see if that could possibly make any money...
It really didn't occur to me that the AI haters posting on this website were unaware that Grok is foundational to "The Algorithm" and that every post they make has to be analyzed by Grok. Get thee to Bluesky, or you're part of the problem.
@ross457547 By posting anything on this website, you cause Grok to be invoked to analyze your post. Because you've replied more than me in this conversation, you've caused Grok to run more than I have and you've wasted more water and caused more heating than me. There's no avoiding Grok on X
@ross457547 By posting anything on this website, you cause Grok to be invoked to analyze your post. Because you've replied more than me in this conversation, you've caused Grok to run more than I have and you've wasted more water and caused more heating than me. There's no avoiding Grok on X