The tech sphere has recently been buzzing about the announcement of the NEO Home Robot. For $20,000 or a monthly subscription fee of $499, you can pre-order your friendly humanoid home assistant marketed to do your chores: dusting and ordering your bookshelves, vacuum cleaning, bringing you flowers, carrying your groceries from the car, emptying the dishwasher, and probably a lot more. It also can act as a social robot, telling you jokes, giving you personal advice, acting as a knowledge store, and more. The marketing is certainly brilliant: who doesn’t want to outsource these daily, mundane tasks to an artificial intelligence covered in a soft and cozy shell with a friendly, yet neutral face?

I certainly do! But then I have to think about Cory Doctorow’s concept of enshittification, the process describing the rise and decline in quality and utility of Silicon Valley internet platforms. As much as I love robots, I don’t have faith in Silicon Valley.

Enshittification

What is ensittification? First, new digital platforms are launched with a high-quality service, offering a truly unique value proposition, such as replacing your inefficient, linear TV with HD shows and movies you really want to see when you wish to see them (Netflix). Another example is having access to the entire worldwide music library (Spotify) without the need to buy expensive CDs that just collect dust. In this initial stage, these platforms are cheap or even free to attract as many users as possible.

As the user base of these platforms grows, their CEOs and managers begin monetization, and their focus shifts to prioritize business clients such as advertisers over their users. During this monetization phase, user experiences start to degrade as ads start to show up on your streaming services (Amazon Prime Video 2025) or social media feeds (Facebook around 2012). Samsung Smart Fridges now show you advertisements in your kitchen that you never asked for. Companies start to monetize you by introducing subscription fees or selling your data to dubious ad networks. Another example is the turn away from human-generated user feeds on social media platforms towards algorithmic manipulation and exploitation of user interests to steer their behavior (Facebook, Instagram around 2015).

In the last stage (the self-centric one), the platform deteriorates for both users and business clients as it aims to maximize profits for shareholders while neglecting the initial utility function that made the platform appealing to its users. This is when prices of subscription services hike (Microsoft Game Pass 2025), and users are exploited to the maximum, for example, by paying for new features (LinkedIn Premium Tier) or paying for ad-free YouTube. Zoom famously increased the cost of its premium tier while limiting free meeting durations (its original value proposition during the pandemic), pushing users into the paid plans. Another sign of the third stage is deterioration of content: all social media networks are flooded with cheap AI-generated content (AI slop) and maximum commodization: Instagram and TikTok are reworked as e-commerce platforms aimed at selling you stuff. Another example is Amazon, which now pushes its own private label products over third-party sellers, often uses algorithms that disadvantage sellers, and raises fees, squeezing both vendors and consumers. Another sign of this stage is user lock-in: these platforms crack down on interoperability (such as exporting your Spotify music library) or sharing your passwords with friends and family (as Netflix famously advertised it would not do and then did).

Enshittification of humanoid robots

With this enshittification, we can speculate about what the future of these shiny new Silicon Valley robots will look like:

Initial Stage (Good to Users)

• Robots would launch with a focus on exceptional usability, functionality, and novel features to attract early adopters and build a user base.
• Companies might offer extensive free support, updates, and compatibility to make robots indispensable helpers in homes or workplaces.

Monetization Stage (Good for Business Clients):

• As adoption grows, companies could start monetizing third-party developers and service providers for access to the robot ecosystem. Want to synchronize your robot with your smart dishwasher? Pay for the API access! Your robot wants to access the smart doorbell; there is an app for that (paid). Want to let your robot manage your smart thermostat or fridge? Pay again.
• They may introduce paid app stores, proprietary accessories, or premium features requiring subscriptions or one-time fees. Want to have this new voice model or this new quirky behavior model mimicking a celebrity? Please pay! You want to have access to the new software features, such as language support for Scottish or Saxon dialects (with which voice recognition systems struggle)? Well, well, it costs extra!
• Proprietary accessories will be huge: we just launched a new mop specifically designed for your robot, or your robot only works with the mop or vacuum by company X, with which we have a corporate deal. Of course, this is a mop designed to click into the proprietary port, so you can’t really use it with human hands! Each robot brand will, of course, have its own proprietary connector (much like your phone chargers of yore or early-day electric vehicles).
• They will introduce cloud-based subscriptions for new features (think Apple iCloud+), resulting in the need to be always online (if not already the case). The internal storage of your robot is full? Upgrade to this cloud-storage option (and click to agree to full data access)! Complex tasks likely can only be computed in the cloud anyway. Let’s hope you have good internet coverage in your area! Of course this cloud data resides purely in US data centers with government access to the cloud and the Patriot Act.
• Of course the robot will collect data on everything you do, and it does for you: what toothbrush you use and how often. It will analyze what food you buy while sorting it into your fridge and then sell this information to retailers near you and inform you about special offers. It will know if you are a drinker, smoker, or drug addict and when you go to the bathroom. Furthermore, it will know your moods and suggest you take a drink (“from this sponsored brand”) when you seem stressed (hyper-personalized targeting). Of course, your insurance company would love to know about this habit, and the company is willingly selling this data to the highest bidder. It will have access to your credit card and information once it has the functionality of buying things for you. It might buy you a gift to brighten your mood (“from our sponsor!”). With eyes and ears in your home, having access to your most intimate things that even smartphones cannot access nowadays, it will know about your quirks, dirty laundry, nasty habits, and secrets you keep under your bed. It would be a shame if criminal hackers or foreign spy agencies hacked your always-online robot and demanded a hefty ransom from you or otherwise released your dirty secrets to the world!

Self-Centric Stage (Abuse of Users and Partners):

• Users may face aggressive upselling of features or services previously included or free, such as maintenance, diagnostics, voice assistant upgrades, or security patches. Of course, your robot will only get security updates for five years or so, and beyond that, it is not supported. With access to your home, smart devices, and payment card data, this becomes a huge cybersecurity risk. Insecure robots that roam around the physical world and can be hacked over the internet because they no longer receive software updates—what could go wrong?
• If you want the features of the next robot generation, well, you might need to upgrade to the next-gen Robot Pro, as only it has the specific hardware to run these new shiny tasks and features (think Windows 11 forcing certain hardware requirements on you even though it runs fine on older hardware).
• Robot interfaces could be cluttered with ads or sponsored content, diminishing user experience: “Yes, I will do the dishwasher, after a word from today’s sponsor…” Of course you don’t have access to the proprietary operating system and cannot install an ad blocker or custom software on it. Of course, there will be EULAs and ad and cookie banners to agree with!
• Companies might restrict interoperability and impose proprietary standards, locking users into costly ecosystems. Want to replace your robot with another brand? Good luck trying to export its learned routines, apps, your private data points, and the files and knowledge surrounding you to a new platform. Of course, robots from different brands might not even cooperate with one another (similar to Apple restricting messaging or AirPlay interoperability for a long time).
• AI behaviors may prioritize corporate-defined “engagement” metrics, promoting consumption patterns beneficial to monetized services rather than user well-being: “let’s have another drink by today’s sponsor, Whiskey brand XY.” “You seem depressed; let’s go shopping!” or “Did you know that pharmaceutical company XY has a new antidepressant? Go ask your doctor about it. I can order online if you like!”
• Robot functionality could be deliberately degraded over time without paid upgrades or forced hardware replacements (“planned obsolescence”). With your current robot vacuum, you have to buy new brushes, wheels, and rolls. With your walking robot, you will likely need to replace the rubber coating on its feet, hands, motors, joints, and more. Because costs have to go down, previous high-quality metal parts will be replaced by cheaper plastic components that break more easily! Want to replace the worn-out feet or coffee-stained soft shell of your robot? It will cost you! Of course, the large battery powering your robot is not user-replaceable, so once it runs out, you have to pay a hefty fee.
• As software quality goes down over time, your robot may see inadvertent freezes, odd behaviors, or reboot cycles. It might even brick because of faulty firmware update errors. It might stop working entirely if AWS cloud services have another failure.

Summing things up, we see the need for alternative socio-technical models governing humanoid robots. The current incentive structure, based predominantly on US venture capital and shareholder logic, creates a dynamic of decreasing quality in many products – especially in the long run. This would turn once-exciting human-robot collaboration into a tool of profit extraction rather than empowerment. Since humanoid robots likely have a higher deployment cycle than, say, smartphones or computers, the need for long-term support matters. Meanwhile, the cloud-first model of computing and large language models become even more problematic in the age of walking machines. Here, the EU could come in, pushing not for cloud-centric compute models for humanoid robots but maybe “on premise” or edge-computing models. It might also be worth looking to Japan, which also has experience in this domain. Whatever the way forward, alternatives are needed.