Prompts for our Grok chat assistant and grok bot on X

suddenlybananas · 2025-05-16T12:11:29 1747397489

Hmmm, very little mention of South Africa in these prompts.

mirashii · 2025-05-16T12:25:29 1747398329

https://www.theverge.com/news/668220/grok-white-genocide-sou...

Another lame excuse and some faux transparency to go with it.

jsheard · 2025-05-16T12:38:55 1747399135

> faux transparency

Yep, this isn't the first time they've made this play. Remember when they open sourced the X/Twitter algorithm in the name of transparency-maxxing? Those repos haven't been updated for nearly two years now.

sidibe · 2025-05-16T13:53:01 1747403581

The repo has served its intended purpose: his legion of bizarre fans can now say "it's open source" and brag about his transparency

ryanschaefer · 2025-05-16T13:36:45 1747402605

Call me cynical but this smells like a PR stunt to get people talking about how much controlling the system prompt matters to everyone’s interaction with assistants.

Most people wouldn’t give it a second thought unless something like this happened and it’s now plain to see how quickly they can be (poorly) manipulated.

readthenotes1 · 2025-05-16T12:45:30 1747399530

Other than https://github.com/xai-org/grok-prompts/pull/3

defrost · 2025-05-16T12:17:27 1747397847

Buried deep in the #include Tom Sharpe Riotous Assembly and #include Jeremy Taylor Ag Pleez Daddy et al material.

Squarex · 2025-05-16T12:37:43 1747399063

It's response for their South African debacle. https://x.com/xai/status/1923183620606619649

miohtama · 2025-05-16T15:46:37 1747410397

Regardless of what one thinks about Mr Musk, this is very good move and ethical thing to do and I hope other LLMs would follow the lead

drivingmenuts · 2025-05-16T20:12:27 1747426347

Ethical or unethical doesn't really matter when it's owned and operated by a racist POS ands his collaborators.

Tadpole9181 · 2025-05-16T20:21:05 1747426865

It's also entirely performative garbage. There's nothing that actually proves these are the (only) prompts being used, and if Elon Musk wants Grok to start pushing bullshit again he will do it. But this time it'll be more subtle and they'll point to this "transparency" to explain how it's not possible.

jsight · 2025-05-16T12:59:51 1747400391

This was interesting: "Grok 3's BigBrain mode is not publicly available." - What is "BigBrain"?

thomassmith65 · 2025-05-16T13:47:08 1747403228

It's a private version of Grok that can query the US Social Security database to locate fertile Elon Musk fans with desirable genes. /s

qoez · 2025-05-16T12:21:44 1747398104

I know this is fake but I like that it puts pressure on openai etc to be more transparent.

juujian · 2025-05-16T12:10:57 1747397457

"We need to feign some transparency and release something. Quick, sanitize the instructions and put something on a repo, doesn't need to be exactly what's in production."

qoez · 2025-05-16T12:21:18 1747398078

Similar to what they did to the X/twitter open source algorithm (which hasn't been updated in 2 years)

_1 · 2025-05-16T12:17:27 1747397847

Pretty basic stuff that you'd been in a blog post from 2022.

belter · 2025-05-16T12:32:34 1747398754

Quite a biased dataset you have there...

thrance · 2025-05-16T12:24:50 1747398290

No mention of the "white genocide" story aka why they need to whitewash their system prompt and are releasing this. They were quite efficient in covering it up.

malfist · 2025-05-16T12:33:05 1747398785

Elon claims that was done by an "unauthorized employee"

I believe that as much as I believe it when a politician says it was an "overzealous intern"

tencentshill · 2025-05-16T13:35:29 1747402529

It must have been Adrian Dittmann!

readthenotes1 · 2025-05-16T12:45:50 1747399550

Other than

https://github.com/xai-org/grok-prompts/pull/3

drivingmenuts · 2025-05-16T12:43:10 1747399390

Why are y'all associating with a neo-Nazi wannabe?

（评论） (comments)

（评论）
(comments)