Guide | 🎲 Vivaldi Search Engines Collection 🎲

[Catweazle]

In the first-ever SearchBench AI evaluation, Andi achieved 87% accuracy, outperforming tech giants and well-funded startups alike. Here's why this matters.

SearchBench AI, developed by Y Combinator-backed Talc, is a new kind of benchmark for AI search engines. Unlike traditional static tests, it generates 900 questions in real-time based on current events. This prevents companies from gaming the system by training on known datasets.

https://github.com/Talc-AI/search-bench

[mikeyb2001]

@Catweazle said in Guide | Vivaldi Search Engines Collection :

In the first-ever SearchBench AI evaluation, Andi achieved 87% accuracy, outperforming tech giants and well-funded startups alike. Here's why this matters.

SearchBench AI, developed by Y Combinator-backed Talc, is a new kind of benchmark for AI search engines. Unlike traditional static tests, it generates 900 questions in real-time based on current events. This prevents companies from gaming the system by training on known datasets.

https://github.com/Talc-AI/search-bench

WOW

[mikeyb2001]

@Catweazle any idea if they used the premium versions of the bottom 3 because if a free AI is better than a paid one thats messed up

[Catweazle]

@mikeyb2001, until now no premium version in sight with a similar business model as in Vivaldi, receiving a commission if you buy something using it's searches (also philosophy and privacy very similar to Vivaldi's). In work there is also a Web Clipper (waitlist), I don't know, but I think in the future there will be a Premium version for companies.

Andi was the first search engine using AI with an independent LLM, long before everyone else, focusing mostly on gathering information from trusted sources and not so much on the Chat capability.

Mainly to "understand" what you are looking for and to be able to offer summaries and explanations of content. It doesn't invent anything and it doesn't log your previous questions either. This is possibly why it is so highly reliable, that I can confirm, using it for more than a year as the main search engine.

[mikeyb2001]

@Catweazle i meant accuracy measurement may be against outdated LLMs where if you Pay for Gemini,Chatgpt or perplexity it may be higher Accuracy

[Catweazle]

@mikeyb2001, not necessary, it's not the question of more o less complex LLM, but of the selection of the most trustworth sources. Mor complex LLM are good for chatbots but not so important for search assistants like Andi. It's enough if it "understand" the question and the content of the sources, no needs of a great conversation capability, which Andi don't have, it does what it is made for and this very well.

[Catweazle]

News from Andisearch

https://docs.google.com/forms/d/e/1FAIpQLScSbEvtNVX0fQuuho6zruMZY036PKz7YLDmXp4YrJPVs-nHUg/viewform?pli=1

[Catweazle]

Scira (Formerly MiniPerplx) is an minimalistic AI-powered search engine that combines multiple data sources to provide comprehensive answers. The name ‘Scira’ is derived from the Latin word ‘scire’, meaning ‘to know’ - reflecting our mission to make knowledge accessible and intuitive.

https://scira.app
https://github.com/zaidmukaddam/scira

https://scira.how?q=%s

[mikeyb2001]

@Catweazle said in Guide | Vivaldi Search Engines Collection :

Scira (Formerly MiniPerplx) is an minimalistic AI-powered search engine that combines multiple data sources to provide comprehensive answers. The name ‘Scira’ is derived from the Latin word ‘scire’, meaning ‘to know’ - reflecting our mission to make knowledge accessible and intuitive.

https://scira.app
https://github.com/zaidmukaddam/scira

https://scira.how?q=%s

Is this a replacement for Andi?

Sorry I've been inactive for a while so I totally missed this conversation

[Catweazle]

@mikeyb2001, it's by fare not a replacement, for this it's too basic, it's only another AI search which can be used for free without privacy problems (FOSS)