top of page


In an era in which coverage and exposure is increasingly governed by social media mechanics and dominant editorial voices, a product is needed which creates balanced, informed perspectives of current events by leveraging the full spectrum of local to international reporting. 

A reader’s understanding of news events is no better than the sources selected to inform them

In the spectrum between human and computation decision-making, or local and international coverage, there is an opportunity to incorporate artificial intelligence and natural language processing to aid the editors which shape coverage and for publications of all sizes to present their reporting on one unified playing field.


1/23 - 5/23




User Experience Design
User Interface Design


Sammy Levin (Personal Project)

Week 4-5: Competitive Analysis

News aggregators are nothing new, but innovators in the field continue to disrupt the way that news is packaged for readers, the way that communities can engage with current events, and the way that journalists and curators are rewarded for their work. 

In order to develop an understanding of the existing product landscape, I compared news aggregators large and small with the intention of understanding how their information architecture, user flows, and feature sets influence the news consumption experience.

This involved trying to integrate each of these aggregators into my daily routine and seeing what habits stick. Personally, Artifact, Google News, and The New York Times are the apps I found myself gravitating towards the most. Artifact recommended me the most articles that interested me, but failed to expose me to broad enough news for me to enjoy using the app. I began to incorporate Google News into my routine when I realized that Artifact was not serving me enough broad coverage for me to be satisfied. Individual publishers' apps such as The New York Times offer a more carefully curated and engaging reading experience, but coverage of a range of topics is limited to a small handful of daily headlines.

It was rewarding to engage in this discovery process because a lot of this research came too little, too late for some of my other projects. For example, I once created a digital service that I later discovered was already largely implemented and I wasn't aware about it. This time around, I feel much more confident in identifying the unique features of each of these competitors and it enables me to build a product that fits into a niche rather than replicate existing features.

Week 6: Perceptual Mapping

On the spectrum from customizability to curation, different news aggregators take starkly different approaches to the way that content is served to users. On one hand, highly customizable services like Substack and Reddit connect readers directly to community curators and writers to build feeds entirely centered around who you follow. On the other hand, services like Apple News and Google News rely on editors to curate top stories of the day. Services like Artifact and Google News also rely on semi-opaque algorithms that serve content based on demonstrated or specified user interests.

On the spectrum from global to community voices, aggregators like Apple News and Artifact rely heavily on thoroughly vetted sources with a global presence. Services like Google News, Ground News, Reddit, and Flipboard serve a balanced mix of community-driven and global stories throughout their feeds. Services like Newsbreak and Substack differentiate themselves by prioritizing sources that cater to niches.

The proposed news app occupies an underserved area in the perceptual map in which the user has moderate control over the content in their feed, and is exposed to a balance between buzz worthy stories on global, national, and local levels.

Week 7-8: Wireframing

News aggregators all package their content in different ways to showcase articles, curated collections, or community engagement with news events. In order to better understand how UI frameworks can influence the way that information is categorized and consumed, I began with a broad exploration of how to represent content as packages for different use-cases.

This work imagines the packaging of a storyline into a fixed size block, which is based off of a previous news app prototype I built called DeepEnd. The DeepEnd app creates a swipe-based interface to explore individual articles across a two dimensional spectrum of breadth and depth for a range of topics. For my next news app, I see the opportunity to build upon this mechanism by packaging multiple stories into one card, and enabling users to curate their own storyline cards. 

At this stage, I frequently revisited the apps that I downloaded for competitive analysis and thought critically about the way that stories were packaged to create narratives. The New York Times for example, will package stories on the home page around particular news events, and have dedicated storyline viewers for ongoing multifaceted issues (such as the SVB/Signature collapse or war in Ukraine). An additional source of inspiration at this stage was Google News' full coverage feature. This feature introduces information like FAQs, tweets, and graphics by weaving them amongst lists of relevant articles. 

Based on what I saw, I created a few conventional packages composed of stories, headlines, subheadlines, leads, and images. I also explored a few ideas I haven't yet seen implemented including freeform image galleries, pull quotes, user comments, and interactive graphics. This would give users the freedom to curate storylines that introduce and engage texts that typically live on different platforms or surfaces.

Weeks 9-10: Information Architecture

This app will feature unique ways to curate and explore content, as well as engage in community discussion. The for you tab features a two-dimensional space to explore the breadth and depth of today's headlines based on the user's personal interests. The local tab organizes stories and pullouts in a linear stack, prioritizing proximity and social engagement. The inbox tab keeps users up to date with relevant content and social engagement. The profile tab stores user information, usage statistics, saved stories, curated storylines, and a subscription manager.

One of the challenges I faced at this stage was designing the information architecture around two drastically different contexts and priorities: for you and local. A lot of the community driven aggregation features that I wanted to highlight in the for you page were challenging to imagine in the context of a local page. For example, extremely community-focused news may not reach large enough of a user base for significant amounts of aggregation to take place.

Furthermore, factors like proximity play a much larger role in shaping what content users want to see in their local are. Stories that are closer to users can be pushed up in the feed and visualizations like mini maps can put issues into context for users. Because community driven aggregation is likely harder to support at the local level, chat and engagement will be on a story by story basis rather than a storyline by storyline basis.

I'm particularly curious to see how the difference between these two feeds will be received at the user testing and critique stage of this project.

Week 11: UI Design

With my information architecture clearly designed and my design system beginning to come together, it was time to move onto the visual design of the UI. I started with the For You page since this is the core surface from which content will be consumed. One of the main challenges at this stage was putting myself in the shoes of a user who would assemble a storyline package from news articles and pieces of information they found.

Another realization I made at this point was that the design system I built out was somewhat limited in terms of the color palette and visual elements. When I tried building my UI to adhere to the color palette I set out originally, it had a bland, filing cabinet feel to it. I checked back on the competitor apps to see how look and feel changed the tone with which news was delivered. I was a fan of Artifact's dark mode palette for its simplicity and typography and clean lines from first party publication apps like The New York Times and the Wall Street Journal. Combining my design system elements with these themes, I built a UI that was intended to be clean, sharp, precise, and straightforward.

Weeks 12-13: UI Design and Prototyping

In order to effectively communicate the interactions and user flow, I implemented Figma's rich suite of prototyping features. I had to return to the drawing board and familiarize myself with tutorials on how to build carousel interactions in order to effectively simulate the behavior of swiping across storyline packages and seeing FAQs and user comments roll past.

Something that I will have to resolve fairly soon is what to do about multidirectional swiping and generally dealing with the complexity of having 10+ storyline packages in the final prototype. It takes quite a few frames to prototype this interaction as intended, which means that clutter builds exponentially as content builds linearly.

Weeks 14-15: Expert Crit

Frame 4.png

In addition to the extensive research that informed my initial design decisions, I was fortunate to have had the opportunity to conduct crit sessions with a few industry-leading experts in news design. These people were able to provide a fresh and informed perspective on the challenges they anticipated in launching this platform, and informed me of existing technologies and solutions that could determine future development.

Here are a few key insights I extracted from these sessions:

Moderation and Guardrails

One of the key concerns posed during these crit sessions was the extent to which users would be moderated on the platform. Because of the community-centric nature of the platform, it becomes essential to have a firm understanding of how different users may leverage the social structure to accomplish their own needs or serve an agenda. One particularly memorable piece of feedback I received from Eben was that users with an agenda are not an edge case if your platform is built to support them as your primary user base. Product designers that build community platforms tend to be conscious of the “1, 9, 90” rule that estimates 1% of platform users are creators, 9% are augmenters, and 90% are consumers. It then becomes critical to build for the majority but be conscious that the 1% may have conflicts of interest that impact the experience for the rest of the user base.

Preflight Nudge

One option suggested by Adam to explore low-touch automated moderation for commenting was to introduce a “preflight nudge”, an emerging moderation technology that warns users if their comment may violate platform guidelines before they post it. This saves readers the embarrassment of having their posted comments removed or disputed and enables readers to self-moderate before posting based on automated feedback from the platform itself. Products like Perspective API are leading the pack with implementing this technology and measured a 40% reduction in toxicity during testing.


One strength of Troubadour highlighted by a few experts was its improved accountability and traceability compared to similar platforms. When quotes are pulled out of context, they can deep link back to their context, headlines always link to the full story, and related articles offer richer perspective on a topic than any single story can. Paul compared Troubadour’s traceability to Wikipedia’s link structure, enabling readers to dive deeper into topics and branch out to related ones.

Platform Relationships

A few experts voiced concern for how this platform can grow in an extensible fashion that fosters a mutually beneficial partnership with its first party publishers. Paul explained that there’s 2 primary strategies that aggregators use when it comes to publisher relationships. In the case of Apple News, Apple has agreed partnerships between all of its partners and pays them commissions based on subscription and readership. On the other hand, platforms like Artifact and Google News establish a direct relationship between readers and publishers with features like an individual subscription manager for each platform. The benefit of Apple’s “brokered” approach is an integrated experience that isn’t hampered by a large number of individual subscriptions. The downside is that this business model may not be sustainable for all publishers, which leads some (like The New York Times) to withdraw or others to surrender significant revenue to the aggregator.

Aggregated Headlines and Storyline Ownership

Another question raised by experts was whether storylines would have aggregated headlines that could be either user generated or machine generated. On one hand, this establishes the visual hierarchy needed to rapidly absorb information about a given package, but on the other hand raises ethical questions of how aggregated headlines offer a neutral perspective on the issue at hand. For aggregators with editorially curated content, a team of trained editors chooses topic names for given storylines, but this platform might place that type of control in the hands of users. Eben also emphasized that packages with clearer visual hierarchy will be more appealing to readers and possibly perform better.


Eben spoke extensively about his experience designing the Facebook News product and recounted Facebook’s approach to determining publisher credibility by paying attention to factors like whether publishers “steal news” from others without fair attribution, whether they correct errors, and whether they deliberately publish misinformation. Compared to its standard social feed, Facebook’s News product more thoroughly vets key voices since credibility is largely assumed by news aggregator readers. He explained that to build an extensible platform, it would be necessary to look more closely at guidelines for the types of people and content that can make it onto Troubadour


Yuri expressed his curiosity about what the onboarding flow for the platform could look like since this can make or break first impressions for so many platforms. He explained that he enjoyed Artifact’s onboarding process by offering more specific topics to follow as you began to indicate your preferences, but also expressed that their topics were narrowly suited for its primarily tech industry user base. In order to provide the personalization algorithm the context it needs to effectively recommend content, an onboarding flow for Troubadour would need to explore a range of topics and ensure that readers interests are met with the right level of specificity.

bottom of page