Introducing Summary.News - The AI news aggregator that summarizes by perspective
I’m excited to soft launch Summary.News, an AI news aggregator that helps news readers know what's really happening by summarizing perspectives such as liberal, mainstream, and conservative. We aim to restore trust in the news with a more holistic view of stories.
Many people were shocked when inflation was not transitory, Kyle Rittenhouse was declared not guilty, a miracle vaccine did not stop spread, and the Ukrainian counter-offensive quickly ground to a halt. A select few newshounds were not surprised by these outcomes, but it takes a lot of time and energy to hunt, peck, and distill through various news sites to triangulate on what’s really happening.
Trust in the news has unfortunately dropped to only 11% for newspapers, which to be clear are primarily mainstream publications and does not include Fox News. Popular news media typically operates from a factual basis but often omits vital facts. Articles about the Supreme Court Dobbs decision will include opinion polls about support for abortion (61% support abortion access). But articles about the Supreme Court Students for Fair Admissions decision often do not include opinion polls about support for affirmative action (only 33% support affirmative action).
At some point, all media converges on reality. Readers who are repeatedly surprised by outcomes like the persistent inflation and the stalled Ukrainian counter-offensive are increasingly losing trust in the media they read on the way to these realities.
It’s a matter of perspectives, summarized by AI
Summary.News offers news summaries from five very different perspective groupings of news sources: Liberal, Mainstream, Conservative, Alternative, and Eastern Bloc. The Alternative perspective includes commentators such as Glenn Greenwald, Matt Taibbi and the Eastern Bloc perspective is composed of Russian and Chinese state media.
Summary.News users by default see articles from all perspectives in their main feed. However, you can subscribe to a subset of perspectives in the main feed and continue to see all available perspective summaries on a news topic. For example, you can subscribe to only the Mainstream perspective, see articles from sources like CNN, New York Times, and the BBC, which are augmented by holistic summaries of the other perspectives.
We have been tuning our AI model to learn what’s essential to each perspective. It’s important to dig deeper than liberals say “Trump bad” and conservatives say “woke bad.” Sometimes, multiple perspectives are saying the same thing, which is a good validation that there is a consensus on the news item. Other times, there are widely different perspectives on an event.
For subjects like technology and sports where there are typically not divergent perspectives, Summary.News shows the most relevant related articles.
AI translation and summarization of non-English news sources
Summary.News uses AI to translate and summarize Russian, Chinese, and Arabic news sources with perspectives rarely exposed outside of their home markets. The Russian and Chinese sources are grouped as the “Eastern Bloc” perspective, often providing a contrarian take to Western news media. They may be right or wrong, but it’s essential to understand what they are thinking and where they are coming from. In the old days (twenty years ago!), Western media would showcase propaganda like the second Gulf War’s “Baghdad Bob,” but now these perspectives are often no longer included.
How it works
Summary.News is a comprehensive news aggregator with AI and machine learning at every step. We use news sites’ RSS feeds to pull content at regular intervals. Every article is summarized using an AI summary generator. We then use AI to group articles together by topic and recency. Within each group of articles, we then summarize the relevant articles from each perspective. For the feed, we gather an “interestingness” signal from an intersection of sites across the Internet. We then use AI to sort the user’s feed by interestingness and the curations and perspectives the user follows.
We then rerun this entire cycle every fifteen minutes, as various news outlets are typically updating articles or writing new articles about a breaking story.
Summary.News’s AI summaries are just that: a summary. We always point to the original article and the collection of articles forming a perspective summary.
Who needs a new news aggregator?
Many people get their news from news aggregators: Google News has 392 million visits per month, and the Yahoo! home page with news has 3.3 billion visits per month. The issue with these giant news aggregators is that they offer limited perspectives and merely present the news rather than summarize it. On social media, Facebook has de-emphasized news, and Twitter/X is essentially a live stream of everything where it takes time for the casual news reader to catch up on the latest news quickly.
Many Democrats are happy with mass media, but younger Democrats, Republicans, and independents are increasingly dissatisfied. This unique inflection point is an opportunity to reinvent news aggregators with AI summaries by perspective.
The Syndichain blockchain
All of the articles pulled by Summary.News are posted onto Syndichain, a Layer 1 blockchain purpose-built to syndicate content using an eventually consistent consensus mechanism. Blockchains like Syndichain are inherently censorship-resistant since they are immutable and achieve consensus across a worldwide network of nodes.
However, every country has its laws and customs for content. Syndichain Node operators in individual countries can additionally filter content for their users, but their actions are isolated to their country, and they can’t affect what happens in other countries. I’m the founder and CEO of InCountry, where we help multinational companies keep regulated data within a country. Syndichain is the inverse of InCountry in that Syndichain helps countries manage what information enters a country.
Thank you to the Summary.News team and the AI community
A big thank you to our team of seven incredible engineers that were able to create a better Google News. I’m continually amazed by what a small, focused team of excellent engineers can accomplish. And a thank you to the creators of BERT, BART, T5, OpenNLP, and the rest of the AI and NLP community for all your contributions, openness, and collaboration. Being part of a wave of technology with such a great community is wonderful. It’s been a while since that’s happened.