a16z Podcast: Beyond CES: Connected Home Devices, Voice, and More

Episode Summary

Executive Summary: Benedict Evans and Steven Sinofsky argue that CES 2018 showed less a clear grand strategy than a messy mix of board-level bets, org-chart behavior, and product experiments. They conclude the smart home will remain fragmented, voice is useful but not a universal platform, and the best products will reduce friction with minimal UI rather than add screens and settings.

Main Topics: Reading CES through strategy vs. org chart (Priority: 5/5): The speakers caution against treating every CES announcement as a board-level strategy signal. Many products reflect tactical initiatives, team ownership, or internal competition rather than existential company direction. Smart home fragmentation and interoperability (Priority: 5/5): They argue the connected home will not coalesce into one unified system. Different devices, countries, replacement cycles, and use cases will keep the market fragmented, with point solutions often making more sense than a single platform. Big companies bundling hardware and ecosystems (Priority: 4/5): Samsung, LG, Sony, and similar firms want to connect many appliances and home devices to lock in customers and extend their ecosystem reach, but they still need to sell products individually through retailers like Best Buy. Voice as a constrained interface, not a universal platform (Priority: 5/5): Voice recognition is improving, but the bigger limitation is not speech-to-text; it is scaling the set of tasks assistants can reliably answer. Voice works best for narrow, remembered commands, not open-ended conversation. UI minimalism and product design (Priority: 5/5): The conversation repeatedly emphasizes that the best smart devices often need less UI, not more. For appliances, simplicity, automation, and a single-go-button model frequently outperform screens and complicated settings. AI/ML as a technology in search of product form (Priority: 3/5): Google and Amazon are experimenting with how machine learning should surface as consumer products—photos, assistant, voice boxes, bots, glasses—but the market is still early and not settled.

Key Arguments: CES reveals company org structure and internal turf battles as much as coherent strategy; multiple teams may independently launch similar products. A startup should not assume a big-company announcement means the entire company is strategically committed to that space. The smart home will likely stay fragmented because device categories, household habits, country norms, and replacement cycles differ too much. A unified home system is attractive in theory, but consumers buy one broken appliance at a time, not a whole ecosystem. Voice recognition improvements are real, but they solve only the speech-recognition part of the problem, not the harder problem of answering arbitrary queries. Alexa-style assistants are effectively IVRs: each supported action has to be explicitly built, so the system does not scale into a general-purpose platform. For many appliances, fewer questions and less UI are better; intelligence should move into the network/device behavior rather than the interface. Devices that solve a recurring daily friction—like kettles, door locks, or automated drapes—create value by being simple and reliable. Cross-platform support (Alexa, Google, HomeKit, etc.) is often a distribution requirement for hardware makers rather than proof of a single dominant ecosystem. The future is likely more messy and incremental than the sci-fi fantasy of one unified interface controlling everything.

Data Points: CES walking distance: about 10 miles each - They separately walked the CES floor and reviewed what mattered afterward. CES press releases: about 300 - They reference the flood of announcements and coverage to distinguish signal from noise. Company size example: 10,000 to 15,000 people - Used to illustrate why multiple groups inside a large company may launch overlapping initiatives. Home replacement cycle: 20-year replacement cycle - Used for fridges to explain why smart-home convergence will be slow. Hotel automation panel: 6 buttons - An example of a simple, reliable control system for lights, drapes, front door/privacy, clean, and thermostat. Alexa skills scale: 5,000 to 10,000 skills - Used to show that even a large skills ecosystem does not solve discoverability or answerability. Voice/device product timing: 18 months ago - Reference to Benedict Evans’s earlier blog post about voice as a developing product category. Supportive query set: roughly 20 questions - The claim that voice assistants can only answer a limited set of built-in questions reliably. Interoperability standard: AC power - Used as the real historical standard that let home devices coexist, rather than a software platform.

Pivotal Quotes: "you have to kind of take a step back and think: well, which of these things are tactical and which of them are strategic? And which of them are existential?" — Steven Sinofsky: On reading CES announcements and distinguishing meaningful company strategy from noise. "the smart home will remain kind of more messy and more complicated" — Steven Sinofsky: On why unified-home fantasies are unlikely to match reality. "The ideal for the washing machine is there's just a go button." — Benedict Evans: On preferring minimal UI and pushing intelligence away from the device interface.

Implications: Builders should focus on specific user problems, not grand platform narratives. In smart home and voice, the winners will be reliable, narrow, low-friction products with minimal UI and realistic expectations about fragmentation.