The days I felt most alive as a journalist were the days I was out of the office, far from a desk doing actual reporting and documenting history.
When I was a reporter with The Associated Press at the start of my career, I'd grab my Reporter's Notebook and head to court hearings or press conferences, rush off to fires and earthquakes, or travel to cover diplomatic negotiations, elections or wars. Being a reporter gave me permission to speak to anyone so I could share their stories with the world. I'd then head back to my desk or wherever I was staying on assignment to write my article, distilling what I thought were the most important points into a few hundred words, and eventually entering my text into the main piece of software that writers use daily: the Content Management System, or CMS.
Most of the raw material I gathered went nowhere. My scrawlings in reporter's notebooks or typed notes, interview audio files, and dozens more photos than the one or two I uploaded to be published — all that stayed in my backpack or on my computer hard drive.
Previously, that unstructured raw material was just noise. My story had its pre-set space in the news report. Editors ranked what was most important from experience and intuition, and those judgments were reflected in the number of words they felt a story merited. Similar calls decided whether a newspaper story made the front page, or a TV piece made the half-hour newscast.
Even when the web granted us unlimited space, we stuck to those old formats. We gave everyone the same packaged product because that's all we were able to do, and most of the primary material went nowhere.
Those constraints are gone now. Artificial intelligence and large language models mean we can make sense of all that source material that was previously left on the cutting room floor. That data is actually hugely valuable as unique, new information. And at the same time, the limits on how we create stories are gone. Generative AI can craft a story tailored to a single person and produce millions of those stories at once.
Given this change, making "content" the driving force and output for publishing systems no longer makes sense. What we need now is a new form of software for media companies to orient around: the Context Management System.
From content to context
The word "context" has special meaning for AI, as the information a model takes into account to generate a specific response. The better the context, the better the output. If an AI model has a pool of rich, verified information to work with, the results are dramatically better.
AI labs have been racing to expand what's called the "context window," the amount of information a model can hold in its digital brain at once when generating a response. Top models now handle a million tokens, about 750,000 words, or roughly a thousand typical news stories.
In a Context Management System, journalists would upload all the material they've gathered into a central repository where it can be sorted and tagged automatically. They'd include raw audio from press conferences and interviews, documents from governments and companies, background research from online and real-world sources, along with video and photos from the scene of a news event.
The journalist would flag in the system what they think is more newsworthy, for example, highlighting a significant quote that indicates a key development. A journalist could also create their own analysis in whatever format they prefer: text, audio or video. If there are unanswered questions, those would also go into the context management system to be noted for future follow-ups. Any human insights would be given a higher priority in the system than the raw context.
The journalist no longer spends time packaging a story into a fixed format — the inverted-pyramid article, the 2-minute radio hit. Instead, their role shifts to what matters most: deciding what's newsworthy, verifying what's true, and providing the analysis that no algorithm can. The format is handled by the system. The judgment stays human.
Power to the people
What does this mean for the person on the other end, the one actually reading, watching or listening?
Having a system built on context puts the power in the people's hands. The context becomes the raw material that AI can use to create a unique piece of media just for that person at that time. It can be text, audio or video based on what the user prefers, shaped by their preferences, location and time of day. The story can be customized based on where a person lives, their educational level, or what they know based on past consumption patterns.
The content can also be personalized to an individual's own context drawing from the Context Management System. A report on school building safety will show the data based on where your kids go. City council meetings will lead with the new development going up down the block from you. Stories that concern renters will go to renters, and retiree issues to retirees.
To make this more tangible, I built a rough proof of concept using communications from the New York City mayor's office under Zohran Mamdani, who took office Jan. 1, 2026. The demo reshapes press releases, mayoral statements and agency announcements based on who's reading. A parent sees school news first. A renter sees housing policy. A small business owner sees permitting and tax changes. Same raw material, different stories.
The demo only includes government communications, which also makes its limits a useful illustration of what newsrooms could add. A context system built only on official statements tells you what the administration wants you to know. It doesn't tell you whether the policy is working, who's being left out, or what the press release quietly leaves unsaid. That's the layer only journalism provides — verification, follow-up, on-the-ground reporting and institutional memory. Pull that into the same system alongside the primary sources and the personalized story gets closer to the truth. Leave it out and you have PR with better delivery.
The choice for publishers
For publishers and journalists who ignore this: Don't be surprised when human readers stop coming to your websites and mobile apps. Not because the journalism is bad, but because it's more efficient to send an AI agent to gather what you've published, sift out what's truly relevant to the user's own context, and reassemble it in whatever format works best for them.
The smarter move for publishers is to get ahead of it. Instead of only seeing the finished story artifact as their product, publishers could open up their context databases and offer a verified, continuously updated stream of primary material that any AI could draw from.
All those notebooks in my backpack, the photos that never made it into the content management system — that material was never worthless. We just didn't have the technology to use it. Now we do.