The First Rough Draft of What's Next
Sometimes the best way to try new things is to try new things.
That's why I reserve a portion of my time with Local Angle to work on projects with the potential to accelerate innovation in news organizations. This includes infrastructure, tools, libraries, and even one-off examples of new technologies and how they might be applied to problems in news.
Labs work is grounded in the premise that we have not yet figured out the highest and best use of technologies including (but not limited to) generative AI, and the best way to do so is to build practical products and services and put them out into the world. Building tools and examples to makes innovative work easier makes the industry better off.
The goal is not to pick low-hanging fruit, but to look for opportunities that extend beyond the needs of a single organization, with a focus on medium- and long-term upside over short-term wins.
Build with me
I'm always seeking partners and collaborators, both to co-create these projects and to publish practical prototypes to demonstrate their utility.
The ideal partner is a publisher, individual/entrepreneur, or funder with a novel and interesting idea for how technology might be applied to solve a broad problem in the news industry.
It might be a tool, technical library, storytelling technique or business model experiment. If you have an idea but aren't sure how to make it happen, reach out. If it's the right fit, we'll figure out a way to join forces and make it happen.
Areas of exploration
As of September 2025, areas I am actively exploring include:
- Hyperlocal personalization: How might we use large language models, alongside traditional techniques like multi-stage retrieval, to ensure relevant news reaches audiences at the most local level possible?
- Entity extraction and knowledge graphs: How might we turn news archives into structured repositories of community knowledge — and then use those repositories to create new products and services?
- Journalism embeddings: How might we encode latent structures and norms of newswriting into vectors that can be used to organize and analyze them in meaningful ways? Think: vectors capable of representing common story types, such as “game stories” or “profiles” based only on their content and structure.
- Scraping that scales: How might we adopt and build systems that allow us to more efficiently scrape and organize public documents and other information at scale — enabling us to quickly expand the impact of local civic technology projects?
If any of that sounds interesting, I'd love to hear from you.