There is a Japanese word that has been on my mind this summer: Ma. It is hard to translate. I would describe it as an interval. A pause. The space between things that gives them their shape. The rest between notes that turns sound into music. The empty room in which something can still happen. It is space that is held open on purpose.
What if AI could do exactly that for journalism in the future? Give back the space to think rather than crowding out our thoughts?
I admit it feels counterintuitive. Everything points to media moving in the opposite direction. More efficient. More scaled across devices and surfaces. More tailored to our context and behaviour.
From attention to intention
For a long time, digital media has organised itself around attention. We have learned to measure it, predict it and sell access to it. Feeds, alerts and recommendations compete for every free second. Even when a newsroom's ambitions are much broader, the practical conversation often returns to the same things: how to get someone to open the story, how to keep them there, how to persuade them to come back.
Agents are now moving that conversation towards intention. What is someone trying to achieve? What can we help them do? Answer engines give you the answer, not the page where you yourself may find it on. Personal agents promise to brief you in your preferred tone, at your preferred length, about your preferred topics. There is an obvious appeal in organising media around what people are trying to understand or do. Helping a person understand a bill or prepare for a public decision feels more useful than finding another reason for them to scroll. But a system that serves our intentions can still enclose us. Especially when it starts anticipating them, deciding what we probably want before we have had the chance to think about it ourselves. Media that fits like a second skin.
It’s all very exciting. It’s also slightly suffocating.
Because if we carry the logic through, we could end up with a media environment that fits us so closely there is very little room left to move. Every interest recognised. Every hesitation interpreted. Every gap filled with something the system thinks we will find useful. It might be a very comfortable environment. It might also become quite hard to encounter anything that changes the shape of it.
There is a word for that in machine learning: overfitting. A model overfits when it becomes so tuned to its training data that it struggles with new examples. It can perform very good on yesterday’s tasks and fail on tomorrow’s.
The neuroscientist Erik Hoel has a beautiful hypothesis about this. He suggests that dreaming may help prevent our brains from overfitting. In his account, the strangeness of dreams, their wrong turns and impossible combinations, is the point. It injects noise into a mind that would otherwise become too tuned to the narrow routine of its days. The idea is that the brain breaks its own fit so it can keep generalising to a world it hasn't seen yet.
An overfitted media environment is one that knows you so well it can no longer surprise you. I wonder what that does to us over time. How do we learn to cope with a changing world if so little of what we encounter is unfamiliar?
Making room for judgment
But it doesn’t need to be this way. The same technology that games our attention and nudges us towards a familiar path could be used to give us some of that room back. Protecting the space the attention economy wants to fill, and the agentic delegation wants to close.
The future I would like journalism to help build, over the next three to five years, leaves more space than that. By 2030, I would like us to have products that make reporting much more useful in people's lives, while allowing for surprise, unfinished thoughts and the freedom put it down without being drawn into another interaction. Journalism that helps us navigate a changing world and gives us a chance to notice something we had never thought to ask about.
That means being more ambitious about what stories can do. It also means being more careful about what we optimise for.
Sometimes we have a poorly formed question. Sometimes we change our mind while trying to answer it. And sometimes something matters precisely because we had no idea it existed. The unknown unknowns. You cannot ask for what you do not know you are missing. Journalism has always had a role in bringing those things into view. An investigation can introduce a problem nobody had recognised. A report from another country can shift how we see our own. A good interview can leave us with a question that stays around for days.
This is where the idea of Ma can guide us. It asks what might happen in the interval between what a system offers and what a person makes of it. Where there is room for interpretation, disagreement, curiosity. We can design that interval out of the experience. Or we can choose to protect it.
Make no mistake. I am not pleading to make information unnecessarily difficult to use. A person who needs a clear explanation should get one. A person struggling with language, time or access should not have to perform extra intellectual work to qualify for journalism. There is plenty of friction worth removing. The question is where removing it helps someone, and where it starts to take over the work of forming a judgment.
A story might help me compare two proposals without selecting the one it thinks I should prefer. It might explain what is known while leaving a disputed question visibly unresolved. It might give me a place to begin without immediately arranging the rest of my journey. There is a difference between leaving someone lost and giving them room to find their own way.
The story that acts
I call this idea the story that acts. A story that both has and gives agency. It doesn’t only tell you what happened but can also offer a service grounded in the reporting: help with a calculation, a comparison or a practical next step. Rather than a generic chat box under every article, it begins with what this particular piece of journalism has established and what someone might be able to do with it.
As the story becomes an active participant, it gets better at anticipating questions and helping with next steps. It should never make the reader less involved in deciding where to go. There should be room to refuse its suggestion, follow a different thread, or arrive at a conclusion the newsroom did not expect. If there isn’t, we have built a more capable product but perhaps a less capable relationship.
Imagine a local newsroom in 2030 reporting on plans to redevelop an old industrial site. It has examined the council's documents, spoken to residents, questioned the developer's claims and looked at the promised housing and transport connections. It publishes an article with a clear editorial account of what the proposal would change, who stands to benefit and which questions remain unanswered.
You read it because the site is near your home. Alongside the story is a small offer: if you choose a street, the reporting can help you explore the proposed changes around it. You might look at the new walking route, compare the current plan with the previous version, or ask what evidence supports a claim about traffic. The answers draw on the newsroom's work and show where that work runs out.
Perhaps your first concern is whether the road to your child's school will become busier. While exploring it, you encounter reporting about residents who may be displaced by the development. You had not gone looking for that. The story makes the connection visible because it matters to the public decision, even though it sits outside your original question. You can stay with it. You can also return to your original query. The experience leaves both possibilities open.
If you want to respond to the consultation, the story can help you prepare. It can locate the relevant documents, explain the questions the council is asking, and help turn your concerns into a draft you can edit. It might point out that one of your assumptions is not supported by the plan. But it doesn’t decide your position or manufacture your agreement.
You can then ask to hear when the plan changes in a way that affects the question you were following. For several weeks, you hear nothing. The monitoring continues, but no material change has met the agreed threshold. When a revised proposal does appear, you get an explanation of the difference, with a route back to the wider story. When the consultation ends, the offer to submit a response closes too.
In this example, acting and leaving space belong together. The system takes on some of the work of finding, comparing and following. The reader retains the work of deciding what matters, what they believe and whether they want to participate. The story can bring something new to your attention without asking you to do anything about it.
The same principle could apply to a consumer investigation that helps someone examine their bill, or a transport story that compares the implications of a timetable change. These are relatively modest uses. They are also meaningful. Someone might leave with a better question, a more accurate understanding, or a practical problem resolved.
Not every story needs to turn into a service. There should still be reporting we read because it is beautiful, disturbing, funny or difficult to place. A story about a life very different from our own does not need to help us do something. It can simply enlarge our sense of the world. If we replace ‘attention’ with ‘task completion’ as the universal measure of journalism, we will have found another way to make its purpose too small.
In the journalism of tomorrow, there should also be room for encounters we cannot fully explain in advance. A finite edition can include a story chosen by an editor because it deserves to be seen. A story can invite you to explore without deciding where you should end up. A reader can linger on an image, follow a small detail, or stop halfway through. A thought might surface later, while they are doing something else. Not every moment needs to become a signal used to make the next experience more efficient.
What this asks of newsrooms
All this requires room for hesitation on the other side of the story as well.
To make this possible, newsrooms would need tools that help reporters start with a question before producing an answer. Tools that surface conflicting evidence, identify a missing perspective or let someone try an idea without immediately turning it into an article. Sometimes the most useful intervention might be to leave a fragment alone while the reporter works out why it matters.
For a newsroom, this would involve learning to describe the usefulness and limits of its reporting more precisely. What has this story actually established? Which people or decisions might it affect? What can we responsibly help someone understand, compare or do? What would change our assessment? Those questions are editorial work. They cannot simply be passed to a model after publication.
Journalists would need to become more comfortable working with designers and developers on how their reporting could behave more like services. How do you test whether an adapted explanation still preserves the most important facts? How do you recognise when a helpful suggestion becomes a nudge towards a particular outcome? At what point does this turn journalism into activism?
The responsibility continues after publication. A calculation can become outdated. A correction may affect an answer someone has already received. A newsroom offering these capabilities needs someone who can maintain them and a way to withdraw them when their basis no longer holds. By 2030, this could start with a few stories that offer a useful service, and grow only if the benefit is worth the work of maintaining them.
A new sense of value
I do not have a definitive economic model for this more breathable form of journalism. A calmer interface does not make reporting cheap. An acting story can create new costs as well as new value. And the people who would benefit most from help navigating public decisions may have the least money to spend on it.
Still, there is a different value exchange worth exploring. What might people value enough to support, beyond access to more content?
Being done. A product that respects their time enough to end. An unread pile can bring fatigue and guilt, however good the journalism. A useful ending might make a subscription feel like something that helps rather than another obligation.
Being able to act. Stories that help them do something that matters in their lives, with a newsroom accountable for the reporting and the tools it offers.
Being surprised. A trusted source that regularly makes their world larger, not smaller.
Being able to trust the silence. Knowing a story is still being followed without being disturbed until something that matters has changed.
Start with a ma brief
We do not need to wait for an agent that understands everything about us to explore this future. A simple experiment over a couple of months, built around one continuing issue, is enough to get started: a school merger, an energy transition, a hospital reorganisation, a housing plan. Build one story that can help people understand the issue and, if they choose, do something about it.
But before anyone writes a line of code, before the feature list and before the prompt engineering, write a Ma brief. One page. Four questions.
Where does it end? When does a person know they are done, and how does the product tell them?
Where does it pause? Which moves should wait, ask twice, or suggest sleeping on it?
What does it open? What will it offer that nobody asked for, chosen by whom, and labelled how?
When does it fall silent? What is the threshold below which it says nothing, and how can a person check that the story is still being followed?
Keep the first version small. Offer a few supported ways to explore or apply it. Make it easy to see how your question connects to the wider issue. Test it with a small group that includes people who rarely use the newsroom's products. Compare their experience with the article on its own and the newsroom’s existing alerts.
Did the people testing it understand the issue more accurately? Did they encounter a relevant question they had not considered? Could they take a useful step on their own terms? Could they finish when they had what they needed, without feeling pushed to keep interacting or take a suggested action?
Record the failures and maintenance costs as carefully as the successful uses, then share the findings publicly so other newsrooms can learn what helped, what failed and what it took to maintain.
At the end I would ask everyone who has tried this Ma-experiment a simple question: after spending time with this journalism, do you have more room to think and act for yourself?