Attention devs: the xAI API just got A LOT smarter. With Live Search, Grok can now search through realtime data from 𝕏, the internet, trending news, and more. https://x.com/xai/status/1925244461875175616
I couldn’t find a way to try the live/free demo, seems you have to sign up for the API, buy some credits, and call it yourself. I ran it on a couple prompts & sharing it here for others to see.
See my final thoughts in ⭐️ Overall impression.
Summarize world news
This is the default example from the docs (https://docs.x.ai/docs/guides/live-search)
Provide me a digest of world news in the last 24 hour
You can see the full answer here: World Discourse. It seems like a good summary, nothing too surprising (mostly updates on Israel-Gaza, Ukraine, India-Pakistan).
Summarize memetics news
The most useful thing I think would be being able to notice emerging “persistent”1 trends in specific subcultures or networks2. I gave it the following prompt:
Provide me a digest of the discourse in the last 24 hours about memetics, memetic warfare, propaganda, etc. Especially anything about ‘metamemes’ (these are memes that affect the host’s epistemology, and therefore change the substrate of how all memes spread)
The answer is in Memetics Discourse. It did seem to bring up stuff that I hadn’t seen, which is exciting! Specifically “glyphstorm protocols”
A more abstract discussion on X involves technical or metaphorical language about “memetic compression layers” and “glyphstorm protocols”
This also seems correct as an explanation of how far memetic awareness has spread to mainstream (aka: not yet)
the conversation remains somewhat abstract or niche, with limited mainstream engagement in the sampled data
I re-ran it with "return_citations": True but all of the citations it returned were older than a few months (implying there’s nothing in the last 24 hours?)
Among the tweets it DID flag was Jim O Shaughnessy’s viral memetics thread (which in my book was an important milestone in it going into mainstream awareness):
1/I’ve been think about this a lot recently. Ideas, be they good or bad, often spread like viruses. There is a whole field of thought called memetics, which is the study of memes. Memes are very powerful when effective because they can succeed despite being untrue or illogical https://x.com/jposhaughnessy/status/1057767455446188032
Tpot summary
I asked it for a Tpot Summary but it wasn’t very good. I think it’s missing the “right frame”, like, we’d get much better results if you could constrain it to a specific list off accounts, and see what themes were coming up in their clusters. Kind of like how I visualized my “top trigrams” here (https://x.com/DefenderOfBasic/status/1922110398779494417) so you could see what concepts are recently bubbling up for me, and then in theory if everyone had that, you can aggregate that across the network and do it recursively upwards like that.
Footnotes
-
I think of this as like, a lot of things spike in the discourse and then disappear, but some linger. See Psyop vs Vibe for examples of things that linger, and the different patterns they make ↩
-
this would basically be like prompting an LLM from the future, since LLMs can approximate culture belief, but there’s a time lag between when that belief takes hold in people’s minds, and is reflected enough in the discourse for an LLM to pick it up in the next version ↩