The Intimacy Paradox: Why 'Wrapper' Wearables Are Economically Doomed
- Jude Temianka
- 6 days ago
- 12 min read
The "Loneliness Economy" is projected to be one of the defining market opportunities of the next decade. However, the current wave of hardware solutions—epitomised by the device "Friend"—represents a fundamental failure in business design.
This forensic audit argues that "Friend" is not a product failure, but a business model failure. By analysing the unit economics of Large Language Models (LLMs), the "Drop Culture" go-to-market strategy, and the technical limitations of "wrapper" hardware, I demonstrate why intimacy cannot be sold as a one-off hardware transaction. Quick Links!
The "Solutionism" Trap
The "Appification" of Human Connection
Silicon Valley has a long history of viewing human emotions as engineering problems waiting to be optimised. This mindset, famously coined by critic Evgeny Morozov as "Technological Solutionism," assumes that complex social dilemmas—like the global loneliness epidemic—can be solved by digitally measuring, tracking and correcting our everyday behaviour.
"Friend" is the latest and most egregious example of this fallacy. It attempts to apply a "Software as a Service" (SaaS) mindset to a biological and psychological condition. It treats loneliness not as a structural societal failure, but as a lack of "content" that can be patched with a $129 notification system, $99 for early birders.
The "Novelty" Curve: Learning from BeReal
To understand why "Friend" is destined to fail, we must look at the "Drop Culture" phenomenon it mimics. The device relies on the same viral mechanics that propelled the app BeReal to stardom in 2022.
BeReal promised "authenticity" and "connection" to cure the curated loneliness of Instagram. It peaked at 73.5 million monthly active users, only to collapse to approximately 16 million by 2024—a drop of nearly 78%.
Why? Because novelty is not utility!
Once the initial dopamine hit faded, users realised the "solution" (taking a photo at a random time) did not actually solve the "problem" of feeling disconnected. "Friend" is following this exact trajectory: a massive viral pre-order cycle driven by novelty, with no plan for what happens when the user realises that wearing a plastic necklace hasn't actually cured their isolation.
BeReal Sources: Business of Apps / Online Optimism (AppFigures Data).
The Wearables Hardware Graveyard
The history of wearables suggests that "Friend" is entering a market segment defined by failure.
According to Gartner, 30% of fitness trackers and smartwatches are abandoned by users within just six months. These devices offer tangible, high-value utility (heart rate, GPS, sleep tracking), yet nearly a third of users still find the friction of charging and wearing them too high to sustain.
"Friend" offers significantly less utility than a Fitbit. It has a claimed 15-hour battery life, meaning it requires daily charging—a habit that is the primary friction point for abandonment. If 30% of users quit devices that save their lives (via health monitoring), the abandonment rate for a device that merely asks you to "hang out" and “chat” but responds by text will likely be catastrophic.
Source: Gartner (Personal Technologies Survey).
The "Bandwidth" Mismatch
The most damning evidence against "Friend" comes from the very market it tries to serve: AI companionship. Data from Character.AI reveals that the average user spends two hours (120 minutes) per day interacting with their AI companions. True digital intimacy is "high bandwidth"—it requires long, deep sessions of engagement.
Source: Andreessen Horowitz (a16z) / Character.AI Blog.
This usage pattern breaks the "Friend" hardware model. The device is built for low-context, infrequent interactions (occasional text messages or short voice notes). It cannot physically support 120 minutes of continuous audio processing and Bluetooth transmission without draining its small battery in a fraction of the day (~4 Hours), or incurring massive cloud computing costs.
"Friend" is trying to sell a low-bandwidth solution (a sporadic necklace) to a high-bandwidth need (deep, hours-long connection).
Forensic Analysis of a "Wrapper"
The "R&D" Myth: Anatomy of a Pivot
The marketing narrative surrounding "Friend" suggests a product born from deep psychological research and hardware innovation. The timeline suggests otherwise!
The device is not the result of a multi-year R&D cycle into human companionship; it is a rapid pivot from a failed productivity tool.
As recently as October 2023, founder Avi Schiffmann was promoting a device called "Tab," a productivity wearable designed to transcribe meetings and track workflow. Following a documented "creativity sprint" in January 2024—a tour through Paris, Tokyo, Bali and Berlin—the concept was abruptly rebranded from a workplace assistant to an emotional companion.
By February 2024, the company had purchased the domain friend.com for $1.8M, cementing the pivot. This timeline reveals that the "soul" of the product was swapped in under 8 weeks, pivoting from "B2B efficiency" to "curing loneliness" without altering the fundamental hardware premise.
Source: 404 Media, Jan 2024 / GeekWire, July 2024
The "Clone" Revelation: Meet Omi
The most damning evidence against "Friend’s" innovation claims is the existence of Omi (formerly known as "Friend" by Based Hardware's founder Nik Shevchenko).
Launched on Kickstarter in April 2024, months before Schiffmann’s official "Friend" reveal in July 2024, this open-source device offers nearly identical functionality: an always-listening wearable that captures audio and processes it via a mobile app. Source: Kickstarter, April 2024 / Seeed Studio, May 2024
The technical similarities are not coincidental; they're structural.
Omi, Nik Shevchenko's device, is built on the Seeed Studio XIAO nRF52840 Sense, a widely available, off-the-shelf Bluetooth Low Energy (BLE) chip designed for basic IoT tasks.
Schiffmann’s device shares this "wrapper" architecture: it is effectively a Bluetooth microphone that offloads all intelligence to a tethered smartphone. The hardware itself contains no "AI"—it is a plastic shell for an API call.
There’s much debate as to which founder came up with the idea first.
Was it the Friend turned Omi, or was it the Friend?
Either way, they are irrefutably the same product!
The "Wrapper" Vulnerability
Because both devices (the “Friend/Omi” and “Friend”) lack onboard processing power, they’re technically defined as a "Thin Client Wrapper." They capture audio, compress it, and stream it to a phone, which then uploads it to a cloud provider (likely Anthropic or OpenAI) for processing.
Source: Based Hardware GitHub, 2025 / Nordic Semiconductor Specs, 2024
This architecture creates a fatal business vulnerability.
The company owns neither the Intelligence (the LLM) nor the Infrastructure (the cloud). They own only the plastic shell and the brand. If the underlying AI providers change their API pricing, deprecate their models, or launch a competing feature (e.g., "ChatGPT Voice Mode" on Apple Watch), "Friend" and “Omi” become paperweights overnight.
They’re hardware startups with the defensibility of a drop-shipping business.
The Economics of Boredom
The "Context Creep" Mechanism
To understand why "Friend", “Omi”, and any other wearable wrappers that mimic them are financial time bombs, you must understand how Large Language Models (LLMs) "remember." They don't!
Every time you say "Hey" to the device, the AI doesn't just process that one word. To maintain the illusion of friendship, it re-reads the entire history of your relationship (or a summarised "RAG" of it) to know who you are.
This is Context Re-ingestion.
In the world of LLMs, you pay for what the AI reads (Input Tokens) and what it writes (Output Tokens). As your friendship grows, so does your history. Therefore, the longer you use the product, the more expensive you become to the company.
A loyal friend is a financial liability.
The Math: The "Time to Zero Profit" (TTZP)
Let’s run the unit economics using the current 2025 pricing for "High Intelligence" models like Anthropic’s Claude 3.5 Sonnet or OpenAI’s GPT-4o, which are required to deliver the "empathy" the brand promises.
Revenue:
Retail Price: $99- $129 (One-time purchase, discount/full price).
Hardware Cost (Cogs): ~$55 (Based on comparable Bluetooth IoT BOMs).
Net Margin: $44 (Available for AI compute).
Now, let’s apply the "Character.AI Standard" of usage, where an engaged user spends ~2 hours or roughly 120 interaction turns per day. To simulate a best friend, the model must re-ingest roughly 2,000 tokens of context (your past conversations/personality data) per turn.
Memory:
Daily Input Volume: 120 turns x 2,000 tokens = 240,000 tokens.
Input Cost: ~$3.00 per 1M tokens = $0.72 per day.
Speaking:
Daily Output Volume: 120 turns x 50 tokens (short replies) = 6,000 tokens.
Output Cost: ~$15.00 per 1M tokens = $0.09 per day.
Total Daily Burn: $0.81 per user.
Monthly Burn: $24.30.
Source: Anthropic API Pricing Page / "Introducing Claude 3.5 Sonnet" Announcement, June 2024 / OpenAI API Pricing Page, 2024/2025
The Verdict
At a burn rate of $24.30/month, the $44 hardware margin is incinerated in less than 60 days (1.8 months). If a user actually loves the device and uses it like a real companion, "Friend" loses money by Month 3.
By Month 12, they have lost ~$247 on a single $99 customer.
The Predatory Business Model
This math reveals a dark incentive structure. For the company to survive without a subscription fee, they need you to stop using the device.
The business model relies on "Breakage"—the industry term for unredeemed gift cards or unused gym memberships. They are banking on that earlier Gartner statistic that 30% of wearables are abandoned in 6 months.
If you love the device (Retention), they go bankrupt.
If you get bored and quit (Churn), they keep the $44 margin.
"Friend" is not designed to cure loneliness; it is designed to exploit the "hope" of a cure, collect the hardware margin, and pray you put it in a drawer before the API bills add up. It's a "Subscription-class" service disguised as a "One-time" purchase—an economic anomaly that cannot survive its own success.
Vanity of Valuation
The $1.8M "Asset" that Became a Liability
In February 2024, Avi Schiffmann spent $1.8 million—roughly 72% of his initial funding—to acquire the domain friend.com.
His justification was that the domain itself would provide instant authority and "save money on marketing in the long run," claiming the purchase was "better than a Super Bowl ad".
This decision represents a fundamental misunderstanding of how the modern internet works. A domain is a destination, not a traffic source. Owning the word "Friend" does not mean you own the audience searching for it.
The "Library Billboard" Problem
From an SEO perspective, the purchase is catastrophic. Google’s algorithm explicitly de-prioritises "Exact Match Domains" (EMDs) to prevent spam, meaning the domain itself offers no ranking guarantee.
Worse, the "Search Intent" for the keyword "friend" is transactional zero. Users searching for "friend" are likely looking for the Friends TV show, dictionary definitions, or relationship advice—not a $129 Bluetooth necklace.
Schiffmann effectively bought a billboard inside a public library.
He paid a premium for high footfall, but the audience is there for a completely different reason. He is fighting the Oxford Dictionary and Jennifer Aniston for attention, a battle a hardware startup cannot win.
The $1 Million Contradiction
The most damaging evidence against the "domain as marketing" strategy is the company's own behaviour. If Friend.com were truly a self-sustaining marketing engine, the company would not need to spend heavily on paid acquisition.
Yet, in late 2025, the company launched a massive Out-of-Home (OOH) campaign, spending over $1 million on NYC subway ads. The ads, which plastered 11,000 subway cars with the word "Friend," were largely mocked and vandalised by locals.
This creates a strategic paradox:
The founder spent $1.8M to "save on marketing," then immediately burnt another $1M on ads to explain what the $1.8M domain was for.
The Opportunity Cost of Vanity
For a lean hardware startup, cash is oxygen. Spending $1.8M on a URL before validating product-market fit is not "conviction"; it is negligence.
That $1.8M could have funded:
🔵 36,000 Units of Inventory (at the estimated $50 'Bill of Materials' cost).
👨💻 4 Top-Tier AI Engineers for two years (to build the proprietary memory the device currently lacks).
🛫 36 Months of Runway for a lean team to iterate on the product.
Instead, the capital was frozen in a vanity asset that offers no utility to the user. If the product launch cannot succeed on a burner domain like friend.it.com (cost: $5), or bud.quest ($2), it will not succeed on friend.com.
The Friction of Form Factor
The "Bluetooth Headset" Effect
Social signalling is the invisible killer of wearable tech.
In the mid-2000s, the Bluetooth headset became a universal symbol of "presence disconnect"—it signalled to everyone around you that you were not really there. "Friend" resurrects this stigma but amplifies it. By wearing a glowing pendant that is visibly listening, the user signals that their primary interaction is with a machine, not the room.
We observed this social rejection phenomenon in 2013 with Google Glass. The term "Glasshole" was coined not because the technology was bad, but because it broke the social contract of attention. Research from that era showed that 72% of the public refused to engage with Glass wearers due to the "surveillance gaze".
"Friend" attempts to bypass this by moving the device from the face to the chest, but the fundamental social friction remains: it is an always-on wiretap worn as jewellery.
Source: Toluna, market research study 2014
The "Divorce Necklace": The Privacy Paradox
The device’s core promise—that it "listens to everything"—is also its greatest liability. In a shared social space, consent is mandatory.
While the user may consent to being recorded, their partner, colleagues, and children have not. In an era where 63% of consumers explicitly fear Generative AI compromising their privacy, wearing a device that uploads every ambient conversation to the cloud is an act of social aggression.
Source: Cisco Consumer Privacy Survey, 2024
It creates what I call the "Bystander Friction."
If a user wears this to a dinner party, the guests are no longer having a private conversation; they're generating training data for a third-party corporation. Far from curing loneliness, the device is engineered to repel intimacy. It acts as a "Divorce Necklace"—a wedge that forces those around you to police their speech, ultimately isolating the user further.
The Silence is Structural: Why it is Mute
Perhaps the most telling design choice is that "Friend" communicates via text messages, not audio. "Friend" claims this is less intrusive.
I argue it’s simply physics.
To have a verbal conversation with an AI, the device must perform a "Voice Loop":
Record →↑ Upload → Transcribe → Process (LLM) → Generate Audio (TTS) →↓ Download → Play.
According to 2025 benchmarks for cloud-based voice loops, this chain incurs a latency of 3.1 to 3.9 seconds. This lag is fatal for “connection”.
Human conversation relies on a "turn-taking gap" of approximately 200 milliseconds.
🗣️ < 200ms: Feels like a connection.
🛰️ > 700ms: Feels like a satellite delay.
⛓️💥 > 3000ms: The illusion of presence collapses entirely.
If "Friend" attempted to speak back, the 3-second delay after every sentence would make the interaction feel robotic and transactional. They restricted it to text, not because it is a better experience, but because the hardware physically cannot support the latency required for the illusion of friendship.
The Real Threat – The "Ambient OS" War
The "Elephant in the Room": OpenAI x Jony Ive
While we analyse "Friend," a far more dangerous hardware play is capitalising in the shadows. It's not a Kickstarter clone; it is a collaboration between OpenAI and Jony Ive’s design firm, LoveFrom.
In mid-2025, reports confirmed that OpenAI effectively acquired "io Products," the stealth hardware startup founded by Ive, in a deal valued at over $6 billion.
Unlike "Friend," which is a Bluetooth accessory for a phone, this project aims to be an "Anti-Smartphone." The vision is to dismantle the "App" paradigm entirely. They are betting on "Ambient Computing"—a screenless, intent-based interface where the AI is the Operating System.
The threat to "Friend" is existential.
"Friend" tries to patch AI companionship onto the existing notification overload of 2024.
The OpenAI/Ive device aims to render that entire notification layer obsolete.
If OpenAI controls the Hardware (the "io" device) and the Intelligence (GPT-5/o1), they bypass the Apple/Google duopoly entirely.
"Friend" is bringing a knife to a nuclear standoff.
The Incumbent Defence: The "Enemy of My Enemy"
The incumbents are not sleeping.
Apple and Google are building defensive moats that make standalone AI wearables redundant before they even ship. Apple’s integration of ChatGPT into iOS is a Trojan Horse.
While they currently partner with OpenAI, this relationship is fundamentally doomed. As soon as OpenAI releases its own hardware, it becomes a direct competitor to the iPhone.
My prediction?
Apple will ruthlessly cut OpenAI out of the OS layer once the hardware threat materialises. We are already seeing the groundwork for this: Apple is actively diversifying its AI partnerships, holding talks to integrate Google Gemini and Anthropic’s Claude directly into Siri.
As OpenAI pivots to hardware, Anthropic is emerging as the "neutral" intelligence partner for the incumbents. It’s Switzerland.
Because Anthropic focuses purely on Model Safety and Intelligence—with no hardware ambitions—it is the safer partner for Apple. We are likely witnessing the formation of a new alliance: Apple Hardware + Anthropic Intelligence. This combination delivers everything "Friend" promises (intimacy, safety, context), but it runs on the Apple Watch you already own—a device with a massive install base and deep bio-utility.
The Window has Already Closed
"Friend" is not the future of computing; it is a drift net catching early adopters before the battleships arrive.
The "Intimacy Economy" will not be won by a $129 necklace with a 15-hour battery. It will be won by whoever controls the Operating System of our lives.
In 12 months, we won’t be talking about "Friend." We will be watching Apple and Google fight to keep Jony Ive from making the iPhone obsolete.
Give this "Deep-dive" article a thumbs up if you enjoyed it. Let me know what other companies you'd like me to audit in the comments!