▲ Barbing 6 hours ago | parent | next [-]

  Andrej Karpathy - @karpathy

  Personal update: I've joined Anthropic. I think the next few years at the frontier of LLMs will be especially formative. I am very excited to join the team here and get back to R&D. I remain deeply passionate about education and plan to resume my work on it in time.

  May 19, 2026 · 3:05 PM UTC

▲ Barbing 6 hours ago | parent | prev | next [-]

[flagged]

▲

criddell 6 hours ago | parent | next [-]

For better or worse, that's where his audience is.

▲

skeledrew 5 hours ago | parent | next [-]

I'm pretty sure the audience would follow him wherever he goes.

▲

Barbing 5 hours ago | parent | next [-]

And that’s the point, thank you!

Anthropic should capitalize on this opportunity to undercut their competitor’s platform. They don’t come around often.

	▲	applfanboysbgon 4 hours ago \| parent \| next [-]
		Twitter is not a competitor to Anthropic, some random guy with 2m followers leaving would not undercut it in any way, and xAI is literally partnered with Anthropic and selling their compute now.
	▲	csallen 5 hours ago \| parent \| prev [-]
		I'm sure Anthropic wants him to have as much reach as possible, not abandon his primary distribution channel just to please an extreme minority of the population.

▲

DANmode 5 hours ago | parent | prev [-]

For folks at this scale,

(vs an HN blog author),

networks are communication and dissemination tools,

not principled stances.

▲

Forgeties79 5 hours ago | parent | prev [-]

For worse. It’s distinctly for worse.

	▲	criddell 4 hours ago \| parent [-]
		Eh, as someone who isn't terribly interested in seeing non-stop low effort promotion of garbage AI products, I don't mind that they congregate on a site I rarely visit.

▲

simianwords 6 hours ago | parent | prev | next [-]

The honest reality that people actually like X and use it. They don't make it a point to show off their virtues by boycotting it. Nearly everyone important in the space uses it.

▲

Barbing 5 hours ago | parent | next [-]

I guess they kinda ignore replies and just follow the Karpathy-ies?

Super offending to my sensibilities seeing the extent of slop in replies, and this is months and months ago now. Unbridled poorly prompted GPT-4o replies.

The main posts from smart/funny people are just as good as they would be written elsewhere, yes, but like at a restaurant, atmosphere’s pretty important too… don’t want to eat a tomahawk steak on an airport runway (whether or not the airport’s associated with My Heart Goes Out To You non-Roman non-salutes)

▲

simianwords 5 hours ago | parent [-]

I don't think the slop problem is that high? Maybe its the people I follow. I just mentally filter out obvious slop anyway. The number of "likes" is a good giveaway for whether it is slop or not.

	▲	Barbing 5 hours ago \| parent [-]
		Must’ve seen a thread that was so horrible I nope’d out hard and of course haven’t checked back.[1] Compounding factor must be how I worry about the “nazi bar” accusations, since that story resonated with me (we know the one, polite-talking guy getting kicked out of bar for seemingly no reason). If you can screenshot all the good stuff and put it on Mastodon, thank you! ;) (hehe no perfect solutions to this thang) [1] omg someone prompted GPT to use uncommon words and lowercase letters, and they posted the stupidest model output I’ve ever seen as “their” reaction… it was super disrespectful to make humans read even those few contrived sentences

▲

nggjnvsegbb 5 hours ago | parent | prev [-]

[dead]

▲

philipwhiuk 6 hours ago | parent | prev | next [-]

Anthropic just bought compute off XAI so he's more tightly bound not less.

▲

Barbing 6 hours ago | parent [-]

That’s a gift to both parties there, yes they sure did (ghosttown Grok servers will get some use, Claude customers able to do things such as… using the service)

Karpathy’s so smart and he has to deal with the reply quality we see on XCancel there… one click away from Hacker News and suddenly every reaction is trash instead of insight and deep insight we see here

(Plus the trash I post, but we’ve got some range, not monotonous spam & model output)

	▲	skywhopper 6 hours ago \| parent [-]
		“ghosttown Grok servers will get some use, Claude customers able to do things such as… using the service” … Memphis/Southaven residents will get more air pollution.

▲

etdznots 6 hours ago | parent | prev | next [-]

[dead]

▲

jasonmp85 6 hours ago | parent | prev [-]

[dead]

▲ phillmv 5 hours ago | parent | prev [-]

[flagged]

	▲	CamperBob2 5 hours ago \| parent [-]
		I don't care what their AI is used to generate, as I see that as the user's problem both legally and morally. But I'll be damned if any actions I take make Musk any wealthier, considering what he's doing with what he already has. Nobody who isn't fully on board with white supremacy (et tu, Andrej?) should be using Twitter at this point.