Show HN: Brewly – AI newsletter creator for any topics you want to deep dive in
We are launching Brewly.ai today!
It’s an AI newsletter creator for any topics you want to deep dive in. You can set quality requirements or special instructions for the content, authors or sources you want to see, Brewly will expand your topics into a set of sub-topics/knowledge paths, find and filter the most relevant content and create a personal newsletter for each of the topics.
Here is how it works:
For a given topic, the agent browses the Internet with self-generated keywords based on the chat with you, and checks if any result page is a high-quality article to be added into your newsletter using the following factors: - It finds and summarizes the author information based on LinkedIn and other sources to judge if the authors are credible. For example, an author with the bio “PhD in CS from Stanford University, Principle Researcher at Microsoft Research, Focuses on large-scale pre-training of deep neural networks” is considered credible on the topic “Multimodal LLMs”. - It checks if the article is from a well-known source such as TechCrunch, MIT Technology Review and OpenAI blogs. It also prioritizes articles from official sources. - It evaluates the relevance level of the content to your topic. - It checks how well the article content meets your instructions. - It filters out clickbait articles including flashy titles, promotional contents, poor technical details on a technical subject, etc. - It prioritizes articles that are well structured and have no misspells.
You can ask Brewly to create either a one time result or a daily updated newsletter, chat with the agent to generate more knowledge paths, ask follow up questions, or adjust the focus with new instructions.
The agent will also create a wiki-like report that lays out important knowledge from those selected articles to help you digest, and continue to find new qualified articles(if you choose the daily update option) and update the report correspondingly.
We imagined Brewly to be the AI assistant for us to acquire topic-specific knowledge in a structured and efficient way, to do some heavy lifting in filtering out the best and the newest content from the internet. There's still a long way to go, we would love to hear your feedback and keep improving it!
----------
For a given topic, the agent browses the Internet with self-generated keywords based on the chat with you, and checks if any result page is a high-quality article to be added into your newsletter using the following factors:
• It finds and summarizes the author information based on LinkedIn and other sources to judge if the authors are credible. For example, an author with the bio “PhD in ECE from Duke University, Principle Researcher at Microsoft Research, Focuses on large-scale pre-training of deep neural networks” is considered credible on the topic “Multimodal LLMs”.
• It checks if the article is from a well-known source such as TechCrunch, Forbes and OpenAI blogs. It also prioritizes articles from official sources.
• It evaluates the relevance level of the content to your topic.
• It checks how well the article content meets your instructions.
• It filters out clickbait articles including flashy titles, promotional contents, poor technical details on a technical subject, etc.
• It prioritizes articles that are well structured and have no misspells.
----------
If you have any ideas or tricks on how you filter good content yourself, let us know and we'd love to add them to our agent!