Yaddle is an AI powered answer engine. It offers users quick answers, a distraction-free space, and the ability to verify sources. It's perfect for those who are growing tired of Google's current search experience filled with Ads, sponsored results, bad SEO pages, and general noise.
Yaddle's nuanced understanding of language allows for better results, making it suitable for complex or specialized queries. It's a phenomenal research tool for finding answers, facts, and information.
Fast & accurate
Distraction-free
Sources cited
Follow up questions
Google's mission was to "organize the world's information and make it universally accessible and useful." And they were great at that for decades! But now they add more and more Ads and other distractions. It takes so much time to find what you are looking for.
Yaddle cuts through that noise to deliver you the answer you were looking for. No additional clicking. Just the answer delivered quickly with sources if you want to learn more.
Fast
Clear answer
Search privately
No distractions
Fast
Clear answer
Search privately
No distractions
Yaddle combines the power of traditional web search with the power of generative AI and large language models (LLMs). It uses a technique known as Retrieval Augmented Generation (RAG) to improve the accuracy of a generative AI model. If you're curious about the technical details, check out this blog post. But to keep things simple, here's what happens when you do a search.
An AI language model reads and understands the type of question you are asking. That could be general questions, real-time news, stock quotes, writing essays, or more.
Based on that, it will fetch relevant information to better answer your question. That could be news, real-time stock information, research papers, or high quality 3rd-party sources.
The AI language model will generate your answer using any of the relevant information fetched. By doing that, it improves accuracy and reduces the hallucinations that ChatGPT tend to do.
Yaddle with add citations to the given answer, so you can verify the information yourself. Yaddle will also share great followup questions to allow you to explore further.
The main challenge with LLMs like ChatGPT is that they have knowledge cutoffs (and they occasionally tend to hallucinate). It’s because they’re trained on data up to a specific date (eg Sep 2021). So if you want an answer to an up-to-date question or you simply want to research a topic in detail, you’ll need to augment the answer with relevant sources. This technique is known as RAG (retrieval augmented generation). And in our case, Yaddle simply supplies the LLM up-to-date information from search engines like Google or Bing.
Use Yaddle to get a better, more accurate, and noise-free experience. I'm sure you've noticed the quality of Google search results going down. As they continue to push to make more money, more and more ads and other google services make Google search frustrating. Yaddle gets back to when Google worked, was simple, fast, and you could trust the first result. But Yaddle saves even more time by simplying replying with your answer.
Yaddle leverages traditional web search engine's ability to extract highly relevant content from each source page ("snippets"), which can be fetched very quickly. So Yaddle just has to fetch both search results and call a language model. We are very focused on using fast models and stream back the answers so you can start reading within a second.
Yaddle continues to add the latest and great models available, usually within a day or two of their launch. Today, Yaddle can generate answers using GPT-3.5-turbo, GPT-4, Mistral-7B, Mixtral-8x7B, Mixtral-8x22B, Llama-3-8B, Llama3-70B, Claude 3 Sonnet, Claude 3 Haiku, and Claude 3 Opus. The availability of these models may change depending on usage and cost to the Yaddle team. Learn more about the supported models here
We generally track the quality of each model on the HuggingFace open LLM leaderboard across a number of standard evaluations. We also look at the head-to-head leaderboard using an ELO based way to rank. Today, GPT-4-Turbo and Claude 3 Opus are generally regarded as the "best", though each model has their own strengths. Llama-3-70B is proving to be a worthy competitor to these top of the line models.
The name Yaddle was chosen to evoke the nostalgic days of the older web search engines. A time when web searching was fun, simple, and fast. It's a word that blends names like Yahoo and Google.