What We Do

Belstad is a news platform powered entirely by generative AI. Our concise briefs provide essential information from cross-party perspectives on top news stories, complete with citations and supplementary readings.

Belstad offers a new paradigm for information consumption, optimizing for nonpartisanship, digestibility, and brevity.

How We Do It

1. Event Aggregation

In real-time, Belstad tracks events and collects news articles from an array of sources across the aisle. These articles are then classified based on the political leanings of their sources, using data derived from industry-leading media bias analytics.

2. Curation

Algorithmically, Belstad chooses a subset of the most recent and contextually relevant articles for the event at hand, curating a proportionate spread with regards to source political leaning.

3. Fact Distillation

These articles are then fed to our fine-tuned models, which we’ve trained to cross-reference and distinguish agreed-upon facts, as well as to pick out a balance of partisan perspectives. This is critical, as it provides readers with wholly comprehensive coverage–including perspectives which could otherwise be disregarded. 

4. Results

Because we have also trained our models to optimize for users’ reading experience, what results from this process is a well-flowing briefing that accommodates a traditional news-reading experience.


When choosing sources for political articles, we always aim for an equal proportion of center, left, and right leaning articles. However, events will occasionally see limited coverage from the left or right—a problem outside of any third party’s control. In the absence of source articles from a certain partisanship, centrist sources are substituted in to mitigate potential bias. Additionally, we embed a visible indicator on these briefings to ensure users are aware that the information being conveyed may not equally represent all perspectives.