@iPullRank@dejanseo Dang. So if I’m understanding this right, we can just use Gemma embeddings as a proxy for Gemini embeddings? Or does the fact that they needed rotation to align them mean the out-of-the-box similarity isn’t close enough? Like we meet a rotation matrix or something?
@MParakhin SimilarWeb doesn’t use pixels/cookies. Plus if Edge somehow was able to block all browsing behavior sent to clickstreams, it only has ~5% of browser market share worldwide. Using “severely” here is misleading.
@bryanfcasey Earlier this year on X (can't remember who all did this)… find random library book, take random page & test if quotes from that page was on the indexed web. Most weren’t! I think the point of that is there’s a ton of public knowledge not online or in datasets.
@bryanfcasey I’m def not an expert, but re: 'all public human knowledge' - theyre trained on deliberate/specific datasets, right? Most human knowledge isn't even on the indexed web or in llm training data. What about all the knowledge that’s “public”, but in a physical archive?
Creating content for SEO in a Zero-Click World - A must read analysis and guide by @Chantal_Smink :
* The drawback of SEO content
* Why we all need a new approach
* An alternative approach towards content creation - three main pillars
* The business case for creating zero-click content
Here's the guide: https://t.co/qk2RCerK0d
@JonasSickler Definitely. And I don’t think any content is made JUST for search. Clearly all content has the purpose of being seen/watched/read by a person. To your point, yeah I don’t see how they ever deliver on this message.