I want to share a news storytelling provocation for 2025, coming from a toolmaker who’s been building media prototypes with LLMs over the past year. My hope is that these capabilities will spark ideas from readers working in newsrooms.

First, I’ll share what I’ve noticed in using AI in creative tools and interfaces:

• AI is best used to augment, not replace. Instead of using AI to tell a story, build creative AI tools to help people tell stories.
• Our interfaces will become more welcoming, expressive, and adaptive. This is not just for software, it’s for writing too. Reading is an interface, and we have opportunities to improve the way people get information.

Second, here’s what I’ve noticed in media over the years:

• Everything must be a headline. Anything deeper is often lost or ignored. From a numbers standpoint, engagement drops as news media becomes interactive or subtle.
• Nuanced storytelling is crucial to understanding and navigating our complex world.

We’re caught in this tension. Social media pushes us toward simpler messages — bolder headlines, punchier takes, everything feels flattened. Yet at the same time, the world grows more complex, and stories must have more nuance, more context, and more perspectives.

We can move from flat to dimensional stories with the new interfaces powered by LLMs. I’ve been prototyping what this might look like:

• Zoom Summaries lets you slide between different lengths of the same story. You can have a one-paragraph summary to a two-word headline. The original text becomes liquid, and you can imagine it adapting to its context.
• Summaries Tree renders an article as a tree of nested summaries. Start with the big picture, then go deeper into the parts that matter to you. Explore the story at your own pace, where you want, how you want.
• Website Embeddings Map turns a blog into an explorable map using embeddings, where similar posts cluster together. Hover over any point and get a preview of that post. It’s a new way to get an overview of a site and browse stories.
• AI Reader Assistant augments your reading experience, letting you expand sections inline for more context, get simpler explanations, or dive deeper into concepts — all while staying in the flow of the story.
• SmartNav reimagines how we find and browse information on websites. Instead of hunting through menus and search results, you express what you’re looking for, and the site brings relevant information to you, rendered with a dynamic UI. When tested on MIT’s expansive resources, it surfaced useful content that was previously unsearchable, buried in university blogs and scattered across different pages.

These interfaces offer new ways to think about how we understand complex information. When you let readers explore stories spatially rather than linearly, you tap into the strong human intuition for spatial understanding. When you give them tools to transform text — to simplify it, expand it, or relate it to their own context — they engage more deeply. The LLM text generation in these tools does not replace articles and stories, it augments them. For a familiar example, consider a table of contexts — it doesn’t replace the text, it augments the experience of reading.

The capabilities are here, and now it’s a matter of bringing them to expert storytellers. This coming year, we’ll see more “dimensional” interfaces that let readers explore stories from multiple angles and depths. The bold headline grabs your attention, and then becomes a gateway to deeper exploration and understanding. Now is the time to reimagine what news stories can be when we augment them beyond the constraints of the flat page.