A robot crosses an immaculate living room, a television turns transparent, an artificial intelligence responds without hesitation. At CES in Las Vegas, the future fits into a three-minute demonstration. Real life begins afterward: an unstable connection, a cluttered room, another subscription, a worn-out battery. In September 2026, this contrast deserves more than a dismissal of gadgets. It reveals two distinct disciplines: making technology desirable and making it usable over the long term. Between the two, the decisive innovations are often those that look least impressive on screen.
The show rewards what people can see
CES is a formidable attention machine. To stand out among thousands of exhibitors, companies need a transformation that is immediately understandable: a screen that disappears, a machine that moves on its own, a voice command that triggers several actions. Lower latency on a home network or the availability of a spare part rarely draws the same crowd.
The 2024 edition offered a striking illustration. LG presented its transparent OLED television, while Samsung showed a new version of Ballie, its small rolling robot equipped with a projector. These presentations explored genuine technical possibilities. But their visual appeal said almost nothing about running costs, maintenance or their value in an ordinary home.
This bias is not necessarily deceptive. A trade show also serves to test an idea, attract partners and persuade distributors. The problem begins when the public confuses three stages: a demonstrated feature, a market-ready product and a device that can reasonably be expected to provide several years of service.
The prototype picks its challenges
At a booth, the environment is prepared. The lighting helps the sensors, objects are carefully positioned, the network can be controlled and an operator intervenes if needed. At home, a robot encounters cables, dark rugs, toys and pets. A voice assistant must understand a request spoken from a distance, with the television playing in the background.
Turning a prototype into a product therefore means handling exceptions, not simply succeeding in the main scenario. What does the machine do when it does not understand? Does it stop safely? Does it explain the problem? Can the user easily take back control? Reliability is also measured by how well a device fails. A predictable device inspires more trust than one that is brilliant but temperamental.
Generative AI widens this gap. The Rabbit R1, unveiled at CES 2024, promised a different way to interact with digital services. Reviews of its early versions notably highlighted the gap between that ambition and the features actually available. Beyond this example, delegating a booking or a purchase requires far more than a convincing answer: authentication, confirmation, error handling and data protection.
The real innovation may be an offline mode
A demonstration readily highlights what a device can do when connected. Its everyday usefulness also depends on what it can still do when the cloud becomes inaccessible. A lock must still open, heating must remain adjustable and a light must remain controllable. Local operation may seem mundane, but it reduces dependence on a connection and on a provider’s commercial survival.
In the smart home, Matter illustrates this less photogenic kind of innovation. The standard, backed by Apple, Amazon, Google and Samsung, among others, aims to make interoperability easier. Its gradual rollout since 2022 has not eliminated all differences between platforms: advanced features and support for them can still vary. But avoiding the need to replace equipment when switching ecosystems represents tangible progress.
The useful test therefore goes beyond successful pairing at a booth. It involves changing phones, replacing the router, adding a device from another brand and then checking that the home continues to function. These are ordinary actions, precisely the ones that a spectacular presentation often leaves out of view.
Affordable does not just mean cheaper
The advertised price creates another optical illusion. A device that is affordable to buy can become expensive because of proprietary consumables, cloud storage or a subscription required for certain features. Conversely, a more expensive device may justify its price if it lasts, uses less energy and is easy to repair. But buyers must be able to compare those costs before paying.
Scaling up production adds constraints absent from the prototype. Manufacturers must secure components, limit manufacturing defects, meet safety requirements and organize after-sales service. Impressive technology made in small quantities does not prove that its manufacturer can deliver a consistent product at scale, at the hoped-for price.
When assessing a CES announcement, a few questions are therefore worth more than a lengthy specification sheet:
- Are the features shown included in the advertised price?
- Which operations require Internet access or a subscription?
- How long are updates promised for?
- Who handles repairs, and with what parts?
Repairability starts at the drawing board
A replaceable battery rarely makes a good viral video. Yet it can determine the lifespan of an entire product. The same logic applies to a charging port separate from the main board, accessible screws or a casing that can be taken apart. These decisions are made early: adding a repair service afterward cannot always fix a design that cannot be disassembled.
The European regulatory framework makes these issues harder to sideline. Ecodesign rules applicable since June 2025 to smartphones and slate tablets cover durability, spare-parts availability and updates, among other requirements. The European energy label also introduces information on repairability. This shift does not cover every device exhibited at CES in the same way, but it changes expectations.
Repair must, however, make economic sense. An available part that costs almost as much as a new device is not enough. Usable documentation, acceptable turnaround times and accessible diagnostics are also needed. For a connected product, longevity ultimately depends on software too: a new battery will not save a device whose essential service has shut down.
Change the demonstration, not the spectacle
CES could make these qualities visible. Show a robot in an unprepared room. Cut the Wi-Fi during a presentation. Time a battery replacement. Display support commitments alongside the price. These tests would be less tightly controlled, but more informative. They would also allow dependable companies to distinguish their products through something other than yet another promise.
What next? From September 2026 onward, a reasonable bet is that on-device AI, durability requirements and subscription fatigue will push more manufacturers to emphasize how independently their devices can really operate. This shift is not guaranteed: spectacle will retain its media advantage. To spot useful innovation, we will therefore need to look beyond the applause: at what keeps working without assistance, what costs little to maintain and what can be repaired rather than thrown away.
