Who Controls the Bridges?
Conversational interfaces may replace the grid of apps. The decisive question is whether they will keep the roads to other services open.
Terms with a dotted underline open ChatGPT in a new tab with a ready-made simple explanation prompt.
A recent essay by Paolo De Rosa, written in his personal capacity, asks a question that deserves much more attention: who will control the interface through which we reach the digital world?
His starting point is convincing.
The familiar grid of icons may gradually lose its position as our main way of using technology. Instead of opening an application, finding a menu and learning its logic, we will increasingly express an intention:
Book this trip.
Renew that document.
Find the right insurance.
Pay this invoice.
Arrange an appointment.
The software will still exist. But it may become less visible. An assistant will stand between the user and the services capable of doing the work.
Paolo calls this the mediation layer.
I agree with the diagnosis. But I would add another question.
It is not enough to ask who controls the interface. We must also ask who controls the bridges behind it.
The interface will win before it rules
The next dominant interface will not win because a regulator declares it superior, or because an architect produces the most elegant diagram.
It will win because people like it.
The grid of icons will not disappear through decree. It will recede when asking becomes easier than searching, when conversation becomes more natural than navigation, and when one familiar place can complete work that previously required five applications.
This is where marketing matters in the deepest sense of the word. Not simply advertising, but the ability to understand what people want, remove friction, create habit and make an experience feel obvious.
Consumers will choose the convenient interface first.
Power will arrive afterwards.
Once an assistant knows our preferences, contacts, routines, permissions, history and payment methods, leaving it may feel less like changing software and more like starting a relationship again from zero.
Convenience can become dependency without ever looking like coercion.
Applications will not die
I do not believe that applications are about to vanish.
Their visible front doors may recede, but their capabilities, databases, rules and specialist knowledge will remain. A conversational interface cannot replace the engineering required to operate a bank, a hospital, a transport network or a public administration.
What changes is the route.
Today, I choose an application and then ask it to perform a task.
Tomorrow, I may describe the task first and let an assistant choose the application, service or agent that will perform it.
That choice may be based on quality. It may be based on price. It may reflect a commercial agreement, a ranking rule, a default setting or the interests of the company operating the assistant.
Unlike a search page, the conversation may not show what was considered and discarded.
The service that never appears in the answer can remain fully operational and still become practically invisible.
The real power lies in the bridges
A bridge is the path that lets one system discover, authenticate, call and exchange information with another.
It may be an API, a connector, an open protocol, a commercial agreement or a permission granted by the user.
To control a market, the future gatekeeper may not need to ban competing applications.
It may only need to:
- refuse their connection;
- charge them more than its preferred partners;
- limit the functions they can expose;
- rank them below its own services;
- withhold identity, context or payment capabilities;
- or make their integration too fragile to be trusted.
The door can remain open while the bridge leading to it is removed.
This is subtler than prohibition, and often more effective.
WhatsApp is both the opportunity and the warning
WhatsApp is strategically exceptional because the conversational habit already exists.
People do not need to learn a new interface. Families, customers, companies and public services are already present in the same conversational space.
The phone number also matters.
A phone number is not a complete civil identity. It can be reassigned, stolen, shared or transferred. But it is a practical and globally understood identifier, connected to a telecommunications system that has spent decades solving reachability, interoperability and basic accountability.
That is not a small advantage.
The digital world has many accounts, usernames and passwords, but very few identifiers that work across borders, devices and services with the familiarity of a phone number.
This is why I have never considered the removal of the phone number from messaging to be an obvious improvement. An optional anonymous handle can be useful, but the link between a person, a device and a reachable number gave WhatsApp part of its practical strength.
Meta now has an extraordinary opportunity. It can make WhatsApp an open conversational environment where many services compete to help the user. Or it can turn that distribution advantage into a system that favours only the bridges it controls.
This risk is no longer theoretical. In June 2026, the European Commission ordered Meta to restore free access to WhatsApp for rival general-purpose AI assistants while an antitrust investigation continues. The issue was not whether those assistants existed. It was whether they were allowed to reach users through the channel. The Commission's intervention is a preview of the problem.
The future conflict may not be about who is allowed to build a service. It may be about who is allowed to cross the bridge to the user.
Europe already has part of the answer
This question is not abstract for me.
I became an Estonian e-resident in 2015 because I believed that reliable digital identity would become foundational. A person should be able to prove who they are, or only the specific attribute required, without recreating their identity separately inside every private platform.
Identity cannot stop at the edge of the web. The same person moves constantly between digital and physical situations: signing online, collecting a document, entering a building, confirming an age or qualification, receiving a delivery. A useful identity layer must preserve that continuity while revealing only what is necessary.
Europe is now building part of that foundation through the European Digital Identity Wallet and the wider eIDAS framework.
The wallet is not a conversational assistant, and it should not become one.
But it can become the trust spine beneath many assistants and services:
- reliable identification;
- verified attributes;
- explicit consent;
- electronic signatures;
- controlled delegation;
- and revocable permissions.
This is where Europe should converge rather than multiply incompatible identity schemes.
We do not need a state-owned super-assistant deciding what citizens should use.
We need an attractive public space in which people can safely choose.
I think of it as a pedestrian zone rather than a palace: a lively common area, governed by clear rules, where citizens carry their own identity and permissions, where public and private services can open doors, and where no single landlord decides which streets are allowed to exist.
eIDAS can provide the passport and the trust rules.
It should not own the shops, select the destination or decide which road we take.
What must remain portable
An open conversational ecosystem will require more than a legal right to compete.
It needs technical interoperability that ordinary users can actually exercise.
At minimum:
- Identity must belong to the user. It cannot be trapped inside one channel or assistant.
- Context must be portable. Changing provider should not mean losing years of useful history, preferences and configuration.
- Permissions must be visible and revocable. The user should know which service can access what, for which purpose and for how long.
- Services must expose their capabilities through open interfaces. A smaller provider should not need a private agreement with every dominant assistant.
- Routing must be contestable. Users should be able to choose preferred providers and understand why another one was selected.
- Actions must produce verifiable receipts. When an agent books, signs, pays or changes something, the result should be inspectable and attributable.
- The channel must be replaceable. Moving from WhatsApp to the web, voice or another messenger should not destroy the service relationship.
Without these properties, openness becomes cosmetic.
A user may technically be free to leave while practically unable to take anything useful with them.
The browser we are missing
OpenAI has already moved toward applications that can be called from inside ChatGPT. That is logical and potentially powerful. Apps in ChatGPT turn conversation into a route toward external services.
But the more ambitious project may not be an app directory inside an assistant.
It may be the browser we are missing for the AI era.
The old browser gave users a general client for documents and applications published across the web. It did not remove every gatekeeper, but at its best it preserved a direct route: type an address, follow a link, reach the service.
An AI-era browser would work with intentions rather than pages.
It could let the user choose:
- which model interprets the request;
- which identity and attributes may be used;
- which memory belongs to the user;
- which services are eligible;
- which agent may act;
- and how the action can be audited or reversed.
Such a project could bring together search, applications, agents, identity, payments, documents, communication and automation.
It could solve not two problems with one project, but twenty.
OpenAI could attempt it. Meta, Apple, Google or a European coalition could attempt it too.
But the decisive test would remain the same:
Is it truly a browser, or simply a new closed shopping mall with a conversational entrance?
Build first, declare later
I am not announcing a new platform here.
Ideas of this scale are cheap until real people choose to use them. Architecture diagrams do not create adoption. Principles do not prove usability. A convincing prototype is still only a prototype until it survives contact with habits, constraints and repeated use.
So the right approach is not to sell the bear's skin before killing it.
It is to identify the strategic constraint early, build quietly, test with real users and avoid dependencies that become impossible to remove later.
Demand must lead.
Marketing appeal will decide which interface people try. Product quality will decide which one they keep.
Public rules and open architecture must decide whether that success becomes a service to users or a form of captivity.
Conclusion
The app-store era is not simply ending.
It is being covered by a new layer.
Applications may become less visible. App stores may become infrastructure. Conversation may become the front door.
The next dominant platform will not need to own every service.
It may only need to own the user's intention and the bridges through which that intention reaches the world.
That is why “Who controls the interface?” is the right opening question.
But “Who controls the bridges?” may determine whether users still have a meaningful alternative.
The future should not be one perfect gate.
It should be a city with many doors, many roads and bridges that remain open.
Convenience should be allowed to choose the interface. No one should be allowed to own all the exits.