In August I travelled to San Francisco for the Swissnex Journalism Lab. I left with one takeaway: the future of news is local.
By 2026, agents will browse webpages and feeds to curate our news. We’ll interact with AI through smart glasses. The people I spoke with see a screen-less future as inevitable—new ways of consuming information that require us to prepare now.
“It’s still journalism, even if we won’t recognize it.”
So I spent time prototyping 3D and Gaussians for XR storytelling using 8th Wall, or MapTiler AR. You’ll see these in my upcoming videos—my goal is to let audiences open and interact with the data and maps I build.
But most importantly, I’m building a suite of AI tools for journalists. Let me tell you more.
Scoutpost.ai
The first tool is Scoutpost—because many reporters don’t have time to manually check dozens of websites every day, or the technical skills to scrape data when they need it.
It’s an AI assistant that collects data, monitors sources, and verifies information through conversational requests. Scrape government databases, track social media, query records—capabilities that small newsrooms typically can’t access.
Scoutpost will be available in a couple months. More to follow, but I’ll let you know in the next issue.
The newsroom in 2030
Ubiquitous facts can’t be copyrighted, but unique local data can be. Global stories will be automated and commoditized. The journalists who thrive will be those creating new data—calling sources, building relationships, maintaining accountability so people share information with them.
AI might push a resurgence of local, shoe-leather journalism. Physical presence becomes the differentiator when everything else can be generated.
Research shows precise AI disclaimers build more trust than vague ones. When we disclose exactly how AI processes information, users trust the output more. Show your process, not just results. Be specific about limitations. AI assists, humans verify.
Video stories
My latest video explains how a French billionaire controls 400,000 hectares across Africa through a web of Swiss companies and colonial-era land titles.
I traced palm oil from European shell companies to a village in Cameroon where 3,000 people live trapped by plantations on their own ancestral land. What I found was a system designed to deflect accountability at every level.
This is how modern extraction works.
Watch it here 👇
Reporter’s toolbox
A random list of things that are interesting and worth looking into:
- Picterra.ai – Training custom AI models to detect features in satellite imagery
- Jonathan Soma’s AI for Newsrooms course – Practical frameworks for integrating AI into reporting workflows
- Marimo – reactive Python notebooks that update automatically when data changes
- LM Studio – Running local language models for data analysis without sending information to external servers
- NotebookLM, Google Pinpoint – Organizing and querying large document collections without hallucinations
- Adobe Content Authenticity – Verifying image provenance and detecting manipulation
- Krea.ai – Generating visual assets and prototyping story concepts