The IFA Berlin product floor looked less like a consumer-electronics showcase and more like a modular-blockchain conference with better industrial design. Anker MindBase. Ugreen HomeAgent. LG ThinQ Claw. LinknLink HomeClaw. A row of “AI home hubs” promising local model inference, semantic understanding of daily routines, and enough onboard storage to make a NAS vendor nervous. Price range: €899 to €9,999. These are not accessories. They are architectural statements.
Then the consumer research landed with an almost audible thud. Horowitz Research data covering smart-home owners shows 32% describing setup or daily usage as complicated. 50% want a single dashboard for the devices they already own. 53% are actively searching for better troubleshooting and technical support. IFA’s own consumer work adds another number: 41% now cite privacy as the single largest barrier to adoption.
Read those numbers against the product announcements and the mismatch is not subtle. Users are begging for plumbing. The industry is delivering philosophers. Users want a stable, unified, explainable system for the lights, locks, cameras, and sensors that already fill their homes. Vendors are responding with expensive brains attached to the same fragile body of disconnected gadgetry.
I have spent the past several years tracing failure modes in cryptographic protocols, and the pattern here is unmistakable. This is the Layer-2 playbook transplanted into the living room: enormous investment in a powerful execution environment, almost none in interoperability, auditability, or the boring settlement layer that makes the rest meaningful. Smart homes are not suffering from insufficient intelligence. They are suffering from insufficient structure.
Nobody Is Asking for a Bigger Brain
The smart home already has a classic distributed-systems problem. Multiple vendors. Multiple radios. Zigbee, Z-Wave, Thread, Wi-Fi, Bluetooth, proprietary mesh networks. Every appliance ships with its own application, its own cloud account, its own update policy, and its own idea of what “secure” means. The result is not a home. It is an orphaned network of specialized nodes that refuse to coordinate.
For years, the industry answered with a familiar strategy: add another box. The hub was supposed to fix fragmentation. Then the bridge was supposed to fix the hub. Then the subscription was supposed to fix the bridge. Now the AI hub is supposed to fix everything by being smart enough to paper over what the rest of the stack stubbornly refuses to do.
The fallacy is architectural. You cannot solve an interoperability deficit with more computational intelligence at a single point, because the failure is not located at the point of reasoning. It is located at the edges: between the door lock and the app, between the vendor cloud and the local network, between the Matter thread boundary and the legacy Zigbee device that never received its promised update.
In protocol terms, the industry is adding a powerful sequencer while leaving the consensus mechanism broken. Smart contracts execute. They don’t care how sophisticated the oracle node is if the underlying state transition cannot be verified. The smart home equivalent is a proprietary hub that claims to understand your morning routine while failing to trigger the bathroom light reliably at 7 a.m.
A local large model cannot repair a flaky radio link. Semantic understanding cannot reconcile two vendor clouds that refuse to expose compatible APIs. An expensive NPU cannot explain to a consumer why their brand-new sensor pairing silently fails after a firmware update. Those are interoperability problems. They have always been interoperability problems. They will remain interoperability problems as long as vendors treat the home as an acquisition channel rather than a functioning system.
The Home Has a Ledger Problem
Cryptographic systems taught me to separate two questions: is the system designed correctly, and is the system behaving correctly in production. Most smart-home products address neither. They are designed as closed appliances and monitored through customer-support tickets. There is no tamper-evident record of what a device did, when it did it, and under whose instruction. There is no audit trail for state changes. There is no verifiable proof that a camera processed footage locally rather than shipping it to a vendor cloud.
The absence of verifiability becomes a privacy liability the moment something goes wrong. Eufy is the canonical case. The company made confident claims about local processing and encrypted video. Security researchers found behavior that appeared to contradict those claims, including cloud-based access to footage that users believed was local-only. The Federal Communications Commission eventually added Eufy, along with Roborock and Ecovacs, to its Covered List under rules targeting equipment that poses unacceptable risk to U.S. communications networks.
Set aside the national-security framing for a moment. What matters for consumers is the underlying trust structure. A hardware company tells you that your data never leaves the device. You cannot verify that claim. You cannot inspect the firmware. You cannot force the vendor to prove that the inference actually happened on-device and that the encrypted stream did not include a copy destined for a remote server.
In my corner of the industry, this problem has a name: trusted setup without a public ceremony. In cryptography, we learned long ago that claims of privacy are worthless without proof. Zero-knowledge systems exist precisely because relying on a counterparty’s word is not a security model. Math doesn’t care how reputable the brand appears in a Super Bowl ad. It only cares whether the claim is internally consistent and externally verifiable.
Smart-home vendors are now asking consumers to accept the most intimate data — motion patterns, presence schedules, biometric, voice, video — on the basis of marketing language. The 41% of consumers who list privacy as their primary barrier are not irrational. They have detected, with whatever intuition or news exposure, that the architecture offers promises instead of proofs.
Privacy Is a Verifiability Problem, Not a Storage Problem
The vendors at IFA Berlin appear to believe that privacy is a function of where bytes are stored. Keep the video on a local drive. Keep the model on a local NPU. Keep the semantic index on a local database. Therefore privacy.
This is the same confusion that afflicted early blockchain projects, when teams claimed decentralization because the server happened to be in a different jurisdiction. Locality is not sovereignty. A device in your closet that is controlled by a vendor’s cloud, updated by a vendor’s signing key, and remotely configurable by a vendor’s support system is not local in any meaningful security sense. It is simply another node in the vendor’s network with your floor plan attached.
The stronger frame is verifiability. If a camera claims to process locally, the vendor should be able to produce a signed attestation, ideally from a hardware root of trust, proving that inference and storage remained on the device boundary. If a hub claims that voice recordings are deleted after processing, there should be an auditable log that cannot be rewritten retroactively. If a robot vacuum claims that mapping data stays on-device, the claim should be testable by an independent party.
This is not exotic technology. Remote attestation exists. Signed audit logs exist. Zero-knowledge proofs exist. Trusted execution environments exist, with their own caveats. The building blocks of verifiable local computation have been deployed in enterprise systems for years. What is missing is the demand signal from consumers and the willingness of hardware vendors to architect for auditability rather than for marketing bullet points.
A smart home is, in the end, a database of human behavior. It knows when you wake. When you sleep. When you leave for vacation. Which rooms you actually live in and which rooms you merely walk through. That database deserves better infrastructure than a vendor’s privacy policy.
The FCC Covered List Is a Governance Event Without a Governance Layer
The FCC listing of Eufy, Roborock, and Ecovacs is worth examining as a governance event, because it reveals how the smart-home industry handles existential risk. When trust breaks, the response is not algorithmic. It is regulatory and opaque. A government agency issues a list. Retailers and federal contractors respond. Consumers learn about it through headlines, long after the technical facts have been settled in a security researcher’s report.
Compare that to how mature open-source communities handle trust failures. A vulnerability gets disclosed. A fix is proposed. Maintainers and stakeholders review the patch. The change is signed, published, and distributed. The community can fork the project if governance fails. There is a transparency layer that persists even after the immediate incident.
The centralized smart-home supply chain lacks that layer. A consumer who bought a device from a company added to the Covered List has no obvious recourse. The device still works. The vendor still exists. But the trust anchor has shifted, and the consumer is left holding a piece of hardware whose long-term viability just deteriorated. Nobody promised them an exit path. Nobody explained what happens if the vendor’s cloud is abandoned or the company is forced to shutter U.S. operations.
In crypto terms, this is the equivalent of a major bridge operator being sanctioned while its users hold wrapped assets on the far side. Liquidity is an illusion until it is withdrawn. A smart-home device that depends on the vendor’s cloud for configuration, updates, or basic function has the same structural weakness. It is not self-sovereign. It is a dependent position on another party’s balance sheet.
The industry’s turn to local AI hubs can be interpreted as a clumsy response to this realization. Store more locally. Infer more locally. Reduce the cloud attack surface. But without verifiable guarantees and open protocols, local storage is just soft custody — user-visible, vendor-controlled, and impossible to audit from the outside.
Matter 1.6 Is the Cross-Chain Promise You Have Already Heard
When the connectivity standards alliance introduced Matter, the industry promised a unified application layer. One protocol. One certification process. Devices from different vendors talking to each other without a pile of proprietary bridges. It sounded precisely like the interoperability narratives that blockchain platforms have sold for years.
The reality has been more modest. Matter devices often require specific routers. Thread border routers behave inconsistently across ecosystems. Multi-admin connections work in demonstrations and stumble in real homes. Matter 1.6, the version showcased by the Open Home Foundation at IFA Berlin, is an incremental improvement rather than a revolution. It includes more device types and better energy reporting, but it does not magically dissolve the incentive misalignment at the heart of the industry.
Every vendor wants its app to be the one you open first. Every vendor wants its voice assistant to answer the query. Every vendor wants the telemetry data that makes its own AI more useful. Interoperability, when it happens, is a grudging concession rather than a design principle.
This is not a technical limitation. The technology for device-to-device communication has been mature for years. The limitation is economic. Genuine interoperability commoditizes the vendor’s relationship with the consumer. A fully interoperable smart home is one where the consumer can replace any component without switching all components. That is threatening to the platform business model.
Watch the same pattern in the crypto industry. Cross-chain interoperability has been technically feasible for years. It is economically resisted, because every chain network wants to be the center of settlement. The result is bridge protocols that promise seamlessness and deliver fragmented, risky, and often custodial compromises. Matter 1.6 is a standardization effort fighting the same gravitational forces.
Who Is the Real User? Robots, Apparently
The title of the original analysis that caught my attention was blunt: The smart home industry is building for robots, not people. The evidence at IFA Berlin supports that reading, but the phenomenon is deeper than a product-design critique.
Autonomous agents will increasingly operate in the home. AI agents scheduled by an assistant will want to control the thermostat, the lighting, the entertainment system, the robot vacuum, the cooking appliances. Those agents will need APIs. They will need command authorization. They will need consistent, machine-readable state representations. Designing for agents is a legitimate engineering exercise.
The trouble is that agents are easier to sell to investors than intuitive human interfaces. An agent-native smart home is a narrative of growth: more devices, more data, more automated actions, more recurring intelligence. A human-native smart home is a narrative of restraint: setup once, work reliably, protect privacy, shut up.
Restraint does not move hardware units. AI does. So the industry optimizes for the agent — the entity that will trigger more device sales, more cloud subscriptions, and more training telemetry. Humans are reduced to the entity that occasionally files a support ticket or stares at a dashboard that was designed to serve the machine.
Meanwhile, the actual humans express needs that are almost embarrassingly modest. They want a unified view. They want competent troubleshooting. They want to not feel stupid when a hundred-dollar gadget refuses to join the network. The industry responds with a thousand-dollar box that philosophically understands their kitchen.
There is a lesson from the zero-knowledge world here. A proof system is useful only if the verifier can run it. The most elegant cryptographic construction in the world is worthless if the end user cannot efficiently check the outcome. The smart home has inverted this principle. Vendors are building computationally extravagant systems that no ordinary human can understand, audit, or repair, while neglecting the basic verifiability that would make the system trustworthy.
The Open Home Foundation’s Home Assistant, which supports Matter 1.6 and aggregates thousands of integrations, is the closest thing the industry has to a community-governed alternative. It is not perfect. It requires technical aptitude. It regularly breaks parts that vendors never intended to expose. But it runs on open protocols, keeps local control at the center, and treats the consumer as the principal rather than the product.
Community governance is not a magic solution. Open-source projects have their own failure modes: maintainer burnout, dependency rot, security bugs that linger precisely because nobody is paid to look hard at the code. But open governance has one structural advantage over the proprietary hub model: when trust erodes, the user can leave. The code can be forked. The hardware can be reconfigured. The exit option exists and is exercised by anyone with the skills and time.
The remaining question is whether the smart-home industry will come to terms with its own weakest layer. The plumbing — universal local discovery, portable configuration, signed device reputation, verifiable privacy behavior — is not glamorous. It cannot be demoed with a seductive robot voice. But it is the only layer that can carry the weight of full adoption.
At IFA Berlin 2025, there was no shortage of robots. A mechanical claw descended from a ceiling to retrieve a forgotten package. An AI hub narrated the emotional state of the household. A robot vacuum mapped the apartment with an onboard language model. The exhibition was a catalog of computational ambition.
The data says the users are still waiting for something simpler. Maybe they do not want an architecturally opaque superintelligence in their hallway. Maybe they want the same guarantee that makes cryptographic settlement work: a system whose claims can be checked, whose state transitions can be audited, and whose operators cannot unilaterally rewrite the past.
Until the smart home industry builds that verifiable layer, every AI hub is just another trusted third party, asking users to accept a promise in place of a proof. Smart contracts execute. They don’t pause to admire their own intelligence. Homes, eventually, will demand the same discipline from the machines they shelter.
The industry has two possible futures. In the first, the AI hub becomes a walled-garden sequencer — powerful, convenient, and impossible to leave without replacing everything. In the second, the home adopts an open, auditable, human-verifiable substrate, on top of which both humans and robots can act with permission and accountability. The robots are already here. The question is whether the people will accept a system they cannot inspect, or demand one they can prove.