AI Pins and Pocket AI Devices in 2026: The Post-Humane, Post-Rabbit Reality Check

Direct answer: The "AI pin" category — a standalone wearable meant to replace the smartphone with an ambient AI assistant — went through a complete boom-bust-correction cycle between 2024 and 2026. The Humane AI Pin collapsed into a $116 million fire-sale to HP in February 2025, and Meta shut down the Limitless AI Pendant it had just acquired in December 2025. What replaced the failed "do everything" pitch is a wave of narrower-purpose devices — Plaud's transcription pin (over 1 million units shipped), a rebuilt Rabbit R1 running RabbitOS 2, and cheaper ambient-recording devices like Omi and Bee — that are succeeding precisely because they no longer claim to replace your phone.

A Category That Lived and Died in Public

Few product categories in recent consumer-tech history have gone through as visible a rise-and-fall arc as the AI pin. Unlike most hardware flops, which quietly fade from shelves, the AI pin's collapse played out in front of millions of people, in real time, across YouTube reviews, tech press post-mortems, and — eventually — corporate acquisition filings. The Humane AI Pin launched in 2024 with a founder pedigree (ex-Apple designers), a projector-based interface beamed onto the user's palm, and a pitch that sounded, on paper, like the next computing platform: no screen, no app icons, just a lapel-mounted device you'd talk to and that would talk back, handling calls, messages, questions, and tasks without you ever pulling out a phone.

Reviewers were not kind. Early units ran hot, misheard commands, delivered answers slowly, and burned through battery in a way that undercut the entire "wear it all day" premise. But the deeper problem wasn't a bug list — it was a category-level mismatch between the pitch and what the hardware and AI models of that moment could actually deliver. A device asking users to abandon decades of smartphone muscle memory needed to be dramatically better than a phone at something, not merely different from one. It wasn't. By February 2025, Humane had sold its assets to HP for $116 million — a steep discount from the roughly $850 million-plus the company had reportedly raised and been valued at along the way — and the standalone Humane AI Pin, as a consumer product, effectively ceased to exist.

If Humane's collapse had been an isolated case, it might have read as one startup's execution failure rather than a verdict on the category. Instead, in December 2025, Meta — a company with essentially unlimited capital and its own aggressive wearables roadmap (Ray-Ban smart glasses, Orion prototypes) — acquired Limitless, the maker of the Limitless AI Pendant, and then shut the pendant's consumer service down almost immediately. The Limitless AI Pendant service was confirmed discontinued in the EU, UK, and several other regions following the acquisition, with its underlying technology folded into Meta's broader wearable efforts rather than continued as a standalone product. Two well-funded, well-reviewed-relative-to-Humane attempts at the "ambient AI wearable that replaces your phone" pitch had now both ended the same way: acquired, then killed as a standalone product.

The timing of this whole arc matters too. These devices launched during a period when large language models and agentic AI systems were advancing extremely quickly in capability, and that speed created a genuine temptation for hardware companies to ship a product ahead of what the underlying AI could reliably do — betting that the models would catch up to the marketing before customers noticed the gap. In software, that bet is at least somewhat recoverable: you can push an update, retrain a model, or quietly walk back a feature. In hardware, the bet is much less forgiving. Once a device ships with a projector, a specific chipset, and a battery sized around a particular set of assumptions, you can't meaningfully redesign the physical product to match a slower, more honest capability curve — you're stuck iterating on software within the constraints of hardware that was built for a more ambitious pitch. That mismatch between how fast AI capability was moving in press coverage and demos versus how fast it was moving in reliable, everyday performance is a big part of why this category's first wave landed so hard, and why the correction toward narrower, more achievable promises has proven so much more durable.

Why the "Do Everything" Pitch Kept Failing

It's worth sitting with why this kept happening, because the pattern is instructive well beyond consumer hardware. The AI pin pitch asked customers to make a category-level bet: trust an unproven device, with an unproven interface, running an unproven AI stack, to replace the single most reliable, most capable, most deeply integrated piece of technology most people own. Smartphones aren't just call-and-message machines — they're payment terminals, cameras, navigation systems, authentication devices, and the front door to thousands of apps built over a fifteen-year platform maturation cycle. Asking a $500-$700 pin with a battery that lasts a few hours to functionally replace that isn't an incremental pitch. It's a moonshot.

Moonshots can work, but they need either a dramatically better core experience or a genuinely new capability the incumbent can't touch. The AI pins had neither, at least not yet. Voice-first interaction without a screen sounds elegant in a product demo and frustrating in a grocery store, a loud office, or a video call. Palm-projected displays looked impressive in controlled lighting and nearly unreadable outdoors. And the "Large Action Model" pitch that devices like the original Rabbit R1 leaned on — the idea that the device could learn to operate other apps and services on the user's behalf, essentially automating the smartphone from outside it — ran into the same wall every early agentic-AI product from this period ran into: promising a general capability that the underlying models and integrations weren't mature enough to deliver reliably at launch. Marques Brownlee's widely watched review of the original Rabbit R1 summed up the sentiment across the category: an interesting idea sold as a finished product before it actually was one, with initial functionality effectively limited to a handful of integrations rather than the broad "operate anything" promise in the marketing.

Compounding the product problem was a business-model problem. These devices generally required their own data connectivity, their own cloud AI backend, and often their own subscription — layered costs on top of a phone bill most users were already paying. A device that costs money to buy, money to run, and still requires you to also carry a phone is a very hard sell unless the standalone value is extraordinary. None of the first-generation "do everything" pins cleared that bar.

The Economics Behind the Collapse

Strip away the design language and the keynote-style demos, and both Humane and Limitless were, at their core, capital-intensive hardware businesses trying to bootstrap an entirely new device category against the most entrenched consumer electronics product in history. That's an expensive bet under the best of circumstances. Hardware startups carry costs software startups don't: tooling, manufacturing, supply chain, inventory, returns, and warranty support, all before a single unit generates recurring revenue. Humane reportedly raised roughly $850 million-plus across its funding history and was valued well above that at points along the way — capital that went into R&D, manufacturing setup, and a high-profile launch campaign. Selling for $116 million to HP in February 2025 wasn't just a disappointing outcome; it was a value destruction event on an enormous scale relative to what had been invested, and it happened astonishingly fast relative to the device's public launch.

Part of what made the economics so unforgiving is that these devices asked customers to carry an additional recurring cost — data connectivity and/or a subscription — on top of a phone plan they were, in practice, still paying for, since none of these devices actually let people cancel their phone service. That's a structurally difficult sell: the product needs to justify its own price and its own ongoing cost purely as a value-add, because it isn't actually eliminating any of the costs it was originally pitched as replacing. Contrast that with how the surviving devices in this category price themselves. Plaud and Rabbit's current pricing ($159–$189 and $199 respectively) sit at a fraction of what a mid-range smartphone costs, positioning them clearly as accessories rather than smartphone-tier purchases. Bee's $49.99-plus-$19/month model is even more explicitly framed as a low-commitment add-on. None of the currently successful devices are asking buyers to make a Humane-style bet-the-category purchase; they're asking for an accessory-tier commitment in exchange for a narrow, well-defined benefit.

The Correction: Narrower, Cheaper, Honester

What's notable about 2026's pocket-AI landscape is that the category didn't disappear after Humane and Limitless — it corrected. The devices succeeding now share a common trait: they stopped pretending to replace the smartphone and started being very good at one specific job, usually alongside a phone rather than instead of it.

Plaud is the clearest example. Rather than pitching an ambient assistant for "everything," Plaud built a transcription-and-notes pin: clip it on, press record in a meeting or conversation, and get an accurate, searchable, summarized transcript afterward, with support for 112 languages and pricing in the $159–$189 range. That's a narrow, well-understood job — the same job people used to solve badly with a phone's voice-memo app or a dedicated recorder — done well enough that Plaud has reportedly shipped over 1 million units. There is no pretense of replacing your phone's calling, messaging, navigation, or payment functions. The device does one thing extremely well and gets out of the way otherwise.

Rabbit didn't disappear either — it rebuilt. RabbitOS 2, released in September 2025, was a substantial software overhaul of the original R1 hardware, addressing much of what reviewers criticized at launch: broader and more reliable integrations, a more coherent interaction model, and, notably, a much clearer sense of what the $199 device is actually for rather than an open-ended promise to automate anything. It's a smaller, more honest pitch than the original "Large Action Model will run your digital life" framing, and it's landing better as a result — reviewers have directly compared the rebuilt R1 favorably against the now-defunct Humane Pin.

Below Plaud and Rabbit sit a cheaper, more experimental tier: Omi and Bee. These are ambient-recording wearables built around always-listening capture, transcription, and AI-generated summaries or action items, priced to be an impulse purchase or a productivity experiment rather than a smartphone replacement. Bee combines a roughly $49.99 device with an approximately $19/month subscription for ongoing AI processing. Omi has leaned into developer and community customization — reportedly supporting 250-plus community integrations and options for technically inclined users to bring their own API keys (for providers like OpenAI or Anthropic) rather than being locked entirely into one vendor's cloud stack. That openness has made Omi something of a builder's platform inside the ambient-wearable niche, distinct from Plaud's polished, professional-note-taking focus and Bee's simpler consumer pitch.

Reading the Category Through a Comparisons Lens

If you're trying to decide whether any of this matters to you personally or to your organization, it helps to think of the current pocket-AI landscape less as "AI wearables" as a single category and more as three distinct product types that happen to share a similar physical form factor. The first type is the failed platform play — Humane's Pin and the Limitless Pendant — devices built around the idea that ambient AI plus a wearable form factor could replace or meaningfully reduce smartphone dependence. That type is, for practical purposes, off the table as a 2026 buying decision, since both notable examples no longer exist as supported consumer products. The second type is the professional productivity tool — Plaud being the clearest example — built around one job (transcription and note capture) done reliably and well, sold at an accessory price point, and already validated by real adoption numbers. The third type is the budget ambient-capture experiment — Omi and Bee — cheaper, more consumer-oriented or developer-oriented devices exploring what always-on or on-demand conversation capture is good for beyond formal meetings, without the same maturity or adoption scale as Plaud yet.

Framed that way, the "which AI pin should I buy" question mostly resolves itself once you're honest about which of those three buckets your actual need falls into. Readers comparing options across categories more broadly may find it useful to check Scult's comparisons resources for a structured way to weigh tools against specific use cases rather than marketing claims alone — the same discipline that separates the pocket-AI devices that are actually working in 2026 from the ones that didn't survive their own launch hype.

Who This Actually Affects

It's tempting to read the AI pin story as a curiosity confined to early adopters and tech reviewers, but the underlying dynamics matter to a wider audience for several reasons.

For consumers and professionals, the practical question has shifted from "should I replace my phone with a pin" (a question 2024's marketing wanted you to ask) to "is there a narrow, specific job — meeting notes, conversation capture, hands-free reminders — worth a small dedicated device for." That's a much more answerable question, and it's why the surviving devices in this category look more like specialized accessories (closer to a smartwatch or a dedicated camera) than a platform shift.

For businesses, especially ones in knowledge work, sales, consulting, legal, healthcare, and any other meeting-heavy field, the transcription and ambient-notes sub-category (Plaud, Bee, and similar tools) has real productivity implications — accurate, searchable records of conversations without someone manually typing notes. That utility is genuine and growing, but it arrives bundled with real privacy, consent, and data-governance questions that any organization adopting these devices at scale needs to think through deliberately rather than as an afterthought.

For product and hardware teams — and for any company evaluating whether to build an AI-first physical product — the Humane and Limitless outcomes are a case study worth studying closely. Two well-capitalized teams built genuinely innovative hardware and lost anyway, not primarily because the engineering was bad, but because the product bet (full smartphone replacement, on day one, via voice-only interaction) was mismatched to what the underlying AI and battery/connectivity technology could support at the time. That's a lesson about staging ambition to match technical maturity, not just a lesson about industrial design.

For investors and business strategists, the pattern is a useful signal about how fast hype cycles in physical AI products can invert. Humane raised significant capital on a bold platform thesis and sold for a fraction of its valuation within roughly a year of shipping. Limitless was acquired by one of the best-resourced companies in the world and had its consumer product killed within roughly the same window. Capital and prestige didn't save either product from a fundamental fit problem between promise and delivery.

There's also a talent and technology angle worth noting. Both Humane and Limitless didn't simply vanish when their consumer products were discontinued — their teams and underlying technology were absorbed into larger companies (HP and Meta, respectively). That's a common pattern in hardware-startup failure that's easy to miss if you only track "did the product survive": the underlying R&D, patents, and engineering talent frequently do find a home even when the specific consumer device doesn't. For anyone tracking where ambient-AI and wearable-AI technology ultimately ends up embedded — inside a printer company's future roadmap, inside Meta's smart-glasses line — the Humane and Limitless outcomes are worth watching less as dead ends and more as technology transfers into larger, better-resourced product roadmaps that may resurface these capabilities in different packaging down the line.

The Global Picture

United States. The US is the epicenter of nearly every part of this story. Humane and Limitless are both US companies, and their unraveling — the $116 million HP acquisition in February 2025 and Meta's December 2025 shutdown of the Limitless Pendant — are fundamentally American corporate stories, playing out among Silicon Valley-adjacent hardware startups, US tech press, and US-based acquirers. The narrower successors — Plaud, Rabbit, Omi, and Bee — also treat the US as their primary launch and marketing market.

United Kingdom. The UK has one of the more concrete, verifiable regional data points in this story: the Limitless AI Pendant service was explicitly confirmed as discontinued in the UK following Meta's acquisition, alongside the EU and other regions. For UK-based early adopters who had bought into the Limitless pendant, that meant a functioning device losing its backend service — a tangible, if painful, illustration of the risk buyers take on with acquisition-vulnerable hardware startups.

UAE and Dubai. No distinct regional-specific reporting on this category was found in this research pass. That's not unusual for a still-niche consumer hardware category in its early international rollout phase, but it does mean any claims about AI pin adoption specifically in the UAE market would be speculation rather than grounded fact at this point.

Australia. Similarly, no distinct Australia-specific reporting surfaced in the available coverage. The category's press footprint so far skews heavily toward US and, to a lesser extent, UK/EU markets, with Australian availability and adoption patterns not yet documented in the sources this analysis draws on.

Germany. Germany is captured only within the broader "EU" designation in the Limitless shutdown reporting — meaning German users were affected by the same discontinuation as the rest of the EU, but no Germany-specific detail (adoption numbers, retailer availability, local press reaction) beyond that general EU note has been found.

Europe / France. The same EU-wide Limitless-discontinuation fact applies across the broader European market, France included, per the available sourcing. Beyond that shared data point, no France-specific reporting on AI pins has turned up in this research pass — public discussion of the category in French media appears thin so far relative to US coverage.

China. Interestingly, this sub-category's coverage in the available sources was exclusively US/EU-centric — no distinct China-specific reporting on AI pins or pocket AI devices (Humane, Limitless, Plaud, Rabbit, Omi, or Bee) surfaced in this research pass. That's a notable contrast with the home-robotics and robot-vacuum space, where Chinese manufacturers are frequently cited as leaders; in the AI-pin niche specifically, the visible narrative so far has stayed Western.

A Practical Checklist Before Buying Any Pocket AI Device

Given how quickly this category has produced both genuine successes and high-profile failures, it's worth having a concrete way to evaluate any new pocket-AI device rather than relying on launch-day marketing. A few questions consistently separate the devices that held up from the ones that didn't.

First: what exactly is the device promising, and is that promise narrow or sweeping? Devices that promise one clearly bounded job — transcribe conversations, capture notes, provide quick voice answers — have a much better track record than devices promising to replace or subsume an entire existing product category like the smartphone. Second: does the device depend on a subscription or cloud backend to function, and if so, what happens to your purchase if the company behind it is acquired or shuts down? Both Humane and Limitless are direct, recent illustrations of a device becoming functionally useless once its backend service was discontinued — a risk that's structurally built into almost every AI wearable on the market today, not a hypothetical edge case. Third: how mature is the company's track record specifically with this product, as opposed to how compelling the initial demo or keynote looked? Rabbit's own history is instructive here — its original 2024 launch drew heavy criticism, and it took a substantial follow-up release (RabbitOS 2, a year-plus later) to close the gap between promise and delivery. A device with real usage numbers behind it, like Plaud's reported million-plus units shipped, carries a different risk profile than a device that's still mostly demo footage and press coverage.

Fourth, and increasingly relevant as more of these devices ship as always-on or on-demand recording tools: what are the device's actual privacy and consent mechanics, not just its marketing language around privacy. Ambient-recording wearables capture other people's voices as a structural feature, not an edge case, which means the privacy question extends well beyond the buyer to everyone around them. Any organization or individual planning to use one of these devices regularly, especially in professional contexts, should treat that as a governance question worth answering deliberately before adoption, not an afterthought to be handled once complaints or concerns come up. Readers looking for a broader framework on evaluating and adopting new AI-powered tools responsibly can find related guidance in Scult's resources hub.

What This Means Going Forward

The AI pin category's near-term trajectory looks like specialization, not disappearance. The "replace your smartphone" pitch appears to be functionally dead for now — two attempts by capable, well-funded teams both ended in acquisition-then-shutdown within about a year of each other, and no major player currently seems to be reviving that exact framing. What's replacing it is a set of narrower tools competing on being excellent at one job: capturing and transcribing conversations, generating meeting notes and action items, or providing hands-free reminders and quick AI answers as a companion to (not a replacement for) the phone you already carry.

For businesses evaluating whether any of this matters operationally, the honest answer is: it depends which slice you're looking at. If you're weighing whether to introduce meeting-transcription pins across a sales or consulting team, the productivity case is genuinely strong, but it needs to be paired with a clear policy on consent, recording disclosure, and data retention before rollout — treating it as a workflow and governance decision, not just a gadget purchase. If you're a hardware or product team drawing lessons from Humane and Limitless, the takeaway is about calibrating ambition to what your underlying AI stack can reliably deliver on day one, rather than shipping a platform-level promise your models can't yet back up.

That calibration problem — building an AI-powered product experience that's genuinely reliable rather than an impressive demo that falls apart in daily use — is exactly the kind of engineering discipline that separates products like the surviving pocket-AI devices from the ones that got fire-sold. Teams building their own AI-driven features, whether that's a physical device, an internal tool, or a customer-facing agent, run into the same fundamentals: solid orchestration around the AI, clear scoping of what the system actually promises to do, and rigorous testing against real-world conditions rather than staged demos. Organizations weighing how to build that kind of dependable AI functionality into their own products, rather than shipping something that reads well in a keynote and falls apart in daily use, often benefit from partnering with a team that specializes in production-grade AI agent development and automation — see Scult's AI agents and automation work for an example of what that looks like applied to real business workflows rather than lapel hardware.

The other durable lesson from this cycle is about honesty in product marketing. The devices that are actually working in 2026 — Plaud, the rebuilt Rabbit R1, Omi, Bee — succeed in large part because their pitch matches their delivery. None of them claim to replace a smartphone. All of them are explicit about being a narrow, useful add-on. That restraint, ironically, is what's letting the category survive its own hype cycle instead of being buried by it.

What Businesses and Buyers Are Actually Asking About AI Pins

Can the Humane Pin or Rabbit R1 actually replace a smartphone?

No — and this is the central lesson of the whole category's 2024-2026 arc. The Humane AI Pin was pitched explicitly as a smartphone replacement, built around a screen-free, voice-and-projection interface, and it failed commercially badly enough that Humane sold its assets to HP for $116 million in February 2025. The original Rabbit R1 made a related but distinct promise — that its "Large Action Model" could operate other apps and services on your behalf — and launched with only a handful of working integrations, far short of that pitch. The rebuilt Rabbit R1, running RabbitOS 2 since September 2025, is a better and more capable device, but even it isn't marketed as a full smartphone replacement anymore. Every device still commercially healthy in this category today — Plaud, the current Rabbit R1, Omi, Bee — positions itself as a companion to your phone, not a substitute for one. If a pitch today claims full smartphone replacement, treat that claim with real skepticism given the category's track record.

What are the best Humane AI Pin alternatives that actually work?

The devices that have held up where Humane collapsed share one trait: they picked a narrow job instead of trying to replace the phone entirely. Plaud's transcription pin is the standout, having reportedly shipped over 1 million units on the strength of accurate, 112-language meeting and conversation transcription at a $159–$189 price point. The rebuilt Rabbit R1, now running RabbitOS 2, is a stronger and more honest successor to its own rocky 2024 launch, priced at $199. On the budget end, Omi and Bee both offer ambient-recording and AI-summary functionality at lower price points — Bee pairs a roughly $49.99 device with a $19/month subscription, while Omi has built out a more developer-friendly ecosystem with 250-plus community integrations. None of these claim to replace your phone; they claim to do one job — usually capture and understanding of spoken conversation — better than your phone's built-in tools do.

Which AI wearables actually work in 2026 — Rabbit R1, Plaud, Meta, Omi?

Coverage converging on this question in 2026 generally points to the same short list: Plaud's NotePin for professional transcription, the rebuilt Rabbit R1 (RabbitOS 2) for a broader but still bounded set of AI-assistant tasks, and budget ambient-recording devices Omi and Bee for lighter, cheaper conversation capture. Meta's own entry, the Limitless AI Pendant it acquired, is notably absent from that "actually works" list in 2026, because Meta shut the consumer pendant service down in December 2025 shortly after acquiring it, discontinuing it across the EU, UK, and other regions rather than continuing it as a standalone product. The pattern across the surviving devices is consistent: narrower promises, lower prices relative to Humane's original device, and a clear secondary role alongside a smartphone rather than a platform ambition to replace one. That narrower positioning is functionally what "actually works" means in this category right now.

Did the Rabbit R1 outperform the Humane AI Pin?

In direct comparisons circulating in tech media, yes — reviewers, including a widely discussed Tom's Guide comparison video, found the rebuilt Rabbit R1 outperforming the Humane AI Pin. That comparison is somewhat easier to make now than it would have been at both products' original launches, since Humane's standalone product no longer exists commercially following the February 2025 HP acquisition, while Rabbit continued iterating and shipped a substantially improved RabbitOS 2 update in September 2025. It's worth noting both devices had rough, heavily criticized launches in 2024 — Rabbit's original release drew comparisons to Humane's for being sold as more finished than it was, including limited working integrations at launch. The difference is that Rabbit stayed in the market and fixed much of what critics flagged, while Humane's team and technology were absorbed into HP's roadmap rather than continuing as an independent consumer product line.

Why did the Humane AI Pin fail?

Humane's failure traces to a mismatch between an ambitious platform pitch and what the hardware, battery technology, and AI models could deliver at launch. The Pin asked buyers to replace their smartphone — the single most capable, most reliable device most people own — with a screen-free, voice-and-palm-projection wearable that ran hot, misheard commands, and had limited battery life for an "all-day wear" product. Reviewers were consistently critical of real-world reliability versus the polished product demos. Layered on top of the product experience was a cost problem: the device itself was expensive, and it required its own data plan and subscription, on top of whatever phone bill the buyer was still paying for the phone they hadn't actually replaced. That combination — high price, unresolved reliability issues, and a value proposition that required abandoning a far more capable incumbent device — proved commercially unsustainable, and by February 2025 Humane sold its assets to HP for $116 million, a steep discount to what the company had raised and been valued at.

What happened to the Humane AI Pin after HP bought it?

HP acquired Humane's assets for $116 million in February 2025. In practical terms, that meant the standalone Humane AI Pin, as a consumer product available to buy and use, was discontinued — the acquisition folded Humane's team and underlying technology into HP's own roadmap rather than continuing the Pin as an independent product line in the market. For existing Humane Pin owners, this kind of acquisition-then-discontinuation pattern typically means a service and support wind-down over time, since the product's cloud backend and ongoing updates depended on a company that no longer operates it as a standalone business. This is the same broad pattern that later played out with Meta's acquisition and shutdown of the Limitless AI Pendant in December 2025 — a reminder that with early-stage AI hardware companies, an acquisition by a larger player is not a guarantee the specific product you bought will keep being supported, and can just as easily mean the opposite.

Is the Rabbit R1 worth buying in 2026?

At $199, with the RabbitOS 2 update from September 2025 having addressed much of what reviewers criticized about the original 2024 launch, the Rabbit R1 is a more defensible purchase in 2026 than it was at launch — but it's still worth being clear-eyed about what it is. It's not a smartphone replacement, and reviewers directly comparing it to the now-discontinued Humane AI Pin have generally found it holds up better. If you're drawn to the idea of a dedicated, pocketable AI assistant device separate from your phone — for quick queries, certain automated tasks, or simply as a lower-friction interface than unlocking a phone — the current R1 is a reasonable, low-risk-relative-to-Humane way to try that. If you're expecting it to genuinely replace core smartphone functions, the category's track record (Humane's collapse, Limitless's shutdown) suggests tempering that expectation considerably.

What is RabbitOS 2 and what did it fix from the original Rabbit R1 launch?

RabbitOS 2, released in September 2025, was a major software overhaul for the existing Rabbit R1 hardware, aimed squarely at the criticisms that dogged the device's original 2024 launch — most notably a limited number of working integrations at launch (reviewers noted only a handful actually functioned) against a much broader "Large Action Model can operate any app" marketing promise. The update broadened functional integrations, refined the interaction model, and generally narrowed the gap between what the device claimed to do and what it reliably did in daily use. It's a good example of a hardware company respond to a bad launch not by abandoning the product but by re-scoping its promises to match its actual capability and then rebuilding trust from there — a notably different outcome than Humane's path, where the company sold the product line rather than iterating its way out of the criticism.

How much does the Plaud NotePin cost and what languages does it transcribe?

The Plaud NotePin is priced in the $159–$189 range and supports transcription across 112 languages, positioning it as a genuinely global tool for capturing and transcribing meetings, interviews, lectures, and other spoken conversations rather than a device built narrowly for one language market. That breadth is a meaningful part of its pitch to professional users — consultants, journalists, researchers, and multinational teams — who need reliable transcription across multiple languages without juggling separate tools per language. Combined with its narrow, well-executed focus (record and transcribe, rather than trying to be a general AI assistant), the pricing and language support help explain why Plaud has become one of the clearest commercial successes in the post-Humane pocket-AI landscape, especially relative to devices pitched around broader, less-defined AI-assistant ambitions.

How many Plaud NotePin units have been sold?

Plaud has reportedly shipped over 1 million units of its NotePin, a figure that stands out sharply against the well-publicized failures of Humane and Limitless in the same broad category. That scale of adoption is a strong practical signal that the "narrow, useful job done well" approach to pocket AI devices has real commercial traction, even in a category that generated so much negative press from its earlier, more ambitious entrants. It also suggests a meaningfully sized existing user base for whom AI-assisted transcription and note-taking has become a normal part of their workflow — relevant context for any business evaluating whether to formally support or standardize on this kind of device for meetings, sales calls, or client conversations, given how many individual professionals appear to have already adopted similar tools on their own initiative.

Why did Meta shut down the Limitless AI Pendant?

The specific internal reasoning behind Meta's decision hasn't been detailed in available reporting, but the sequence of events is clear: Meta acquired Limitless, maker of the AI Pendant, and then discontinued the pendant's consumer service in December 2025, shortly after the acquisition closed, folding the underlying technology into its broader wearable efforts rather than continuing it as a standalone product. This mirrors a common pattern in tech acquisitions where a larger company is more interested in a startup's underlying technology, talent, or intellectual property than in continuing to operate its existing consumer product line — sometimes described as an "acqui-hire" dynamic. For a company like Meta that already has an extensive and growing wearables strategy (Ray-Ban smart glasses among them), absorbing Limitless's ambient-AI and pendant technology into that broader roadmap, rather than running a separate standalone pendant product alongside it, is a plausible strategic rationale even without an official public explanation.

Is the Limitless AI Pendant still available in the UK or EU?

No. The Limitless AI Pendant service was confirmed discontinued in the UK, the EU, and several other regions following Meta's acquisition of Limitless in December 2025. For UK and EU consumers who had purchased the pendant, this meant losing access to the backend service that made the hardware functional, since these AI wearables generally depend on cloud processing rather than working as standalone offline devices. This is a concrete, verified example of the broader risk that comes with buying hardware from early-stage AI startups: even a well-funded acquisition by a major tech company is no guarantee that the specific product you bought will continue to be supported — in this case, the acquisition was directly followed by discontinuation rather than continued investment in the consumer product.

What's the difference between the Omi and Bee AI wearables?

Both are ambient-recording wearables built around capturing conversations and generating AI summaries, but they target different users. Bee is the more consumer-friendly, turnkey option: a roughly $49.99 device paired with an approximately $19/month subscription for ongoing AI processing, aimed at buyers who want a simple, low-effort ambient-AI experience without configuration. Omi leans more toward technically inclined users and builders, with a reported 250-plus community integrations and support for bringing your own API key from providers like OpenAI or Anthropic rather than being locked into a single vendor's processing pipeline. In practice, that makes Bee a better fit for someone who wants an appliance-like experience, while Omi suits someone who wants to customize, extend, or self-host parts of the AI pipeline. Both sit below Plaud in price and above it in customization risk — you're trading some polish for flexibility or lower cost, depending on which one you pick.

Do AI pins/pendants record conversations without other people's consent?

This is one of the most important open questions around the entire category, and it doesn't have a single universal answer — it depends on the device, its settings, and the jurisdiction you're in. Ambient-recording wearables like Plaud, Bee, and Omi are, by design, built to capture spoken conversation, which means anyone near the wearer may be recorded without having separately consented, depending on local law and the device owner's practices. Many jurisdictions have two-party or all-party consent laws for audio recording that predate these devices but apply directly to them. Responsible use generally means disclosing that you're wearing a recording device in meetings or conversations, understanding your local consent laws, and choosing settings (where devices offer them) that limit recording to clearly consented contexts. Businesses considering standardizing on these tools for meetings should build explicit disclosure and consent policies before rollout — see Scult's FAQ hub for related guidance on due-diligence questions worth asking before adopting new workplace tools.

Which AI wearable is best for professionals who need meeting notes?

For dedicated meeting-notes use, Plaud is the clearest fit among the devices covered here. It's purpose-built for transcription rather than general AI-assistant tasks, supports 112 languages (useful for multinational teams or client work), is priced at $159–$189, and has already been adopted at scale — reportedly over 1 million units shipped — suggesting a mature, well-tested product rather than an early-stage experiment. Bee is a reasonable lower-cost alternative if the budget case matters more than polish, though its simpler, more consumer-oriented design and subscription model make it less obviously tailored to professional documentation needs. Omi is worth considering specifically if your organization wants control over the AI processing pipeline (via its own API keys and integrations) rather than depending entirely on a vendor's cloud service — a meaningful consideration for firms with strict data-handling requirements around client conversations.

What's the monthly subscription cost for Bee's AI wearable?

Bee's AI wearable pairs a roughly $49.99 device purchase with an approximately $19-per-month subscription that covers the ongoing AI processing — transcription, summarization, and related features — that the hardware alone can't perform locally. This two-part pricing model (low upfront hardware cost, recurring service fee) is common across ambient-AI wearables, since the actual language-model processing generally happens in the cloud rather than on the device itself. When evaluating the total cost of a device like this, it's worth calculating the ongoing subscription cost over a year or more rather than focusing only on the initial hardware price, since the recurring fee is often a larger share of total cost of ownership than the device purchase itself over any reasonably long usage period.

Can I use my own OpenAI or Anthropic API key with an Omi wearable?

Yes — Omi has been reported to support self-hosted or bring-your-own-API-key configurations, allowing more technically inclined users to route the device's AI processing through their own OpenAI or Anthropic account rather than depending entirely on a bundled vendor service. This is paired with a reported ecosystem of 250-plus community integrations, positioning Omi as more of an open, developer-extensible platform than a closed consumer appliance. That flexibility is a meaningful differentiator from more turnkey devices like Bee, and it can matter for organizations or individuals who want more control over exactly which AI provider processes their conversation data, how it's stored, and how it's used — considerations that map directly onto broader data-governance and vendor-lock-in questions companies increasingly weigh when adopting any AI-powered tool, wearable or otherwise.

Is the AI pin/pendant product category dead in 2026?

The "replace your smartphone" version of the category is effectively dead — both major attempts at that specific pitch, Humane's Pin and the Limitless Pendant, ended in acquisition followed by discontinuation within about a year of each other. But the broader category of small, wearable AI-powered devices is not dead at all; it's just repositioned. Plaud's reported million-plus units shipped, the rebuilt and better-reviewed Rabbit R1, and the growth of budget entrants like Omi and Bee all point to a category that's alive and commercially viable once it stopped promising to replace the phone and started promising something narrower and more deliverable — usually accurate conversation capture and summarization. So the honest framing is: the platform-ambition version of this category failed twice publicly; the specialized-accessory version of it is working.

What lesson did AI hardware makers learn from the Humane AI Pin's failure?

The clearest lesson, visible in how the surviving devices in this category now position themselves, is that AI hardware needs to earn trust with a narrow, reliably delivered promise before it can credibly claim a broader one. Humane's Pin (and, in a related way, the original Rabbit R1's launch) tried to skip that step, marketing a platform-level capability — full smartphone replacement, or "operate any app for you" — that the underlying AI models, battery technology, and integration ecosystem of the time couldn't yet support reliably. Every device that's actually working in 2026 does the opposite: Plaud does one job (transcription) extremely well; the rebuilt Rabbit R1 scoped its promises down after a rough launch; Omi and Bee are explicit about being ambient-capture companions, not phone replacements. That's a pattern worth internalizing well beyond hardware — it's a general lesson about not shipping a platform-level AI promise before you can back it up in normal, unscripted daily use.

How does the Rabbit R1's 'Large Action Model' actually work?

The original pitch behind Rabbit's "Large Action Model" (LAM) was that it could learn to operate other apps and services on a user's behalf — essentially watching how a task is performed within an app and then reproducing that action automatically when asked, functioning as a kind of AI-driven automation layer sitting outside the apps themselves. In practice, at the original 2024 launch, this fell well short of the pitch: reviewers found only a small number of integrations actually working reliably, far short of the "operate anything" framing in the marketing. The RabbitOS 2 update in September 2025 broadened and improved on this significantly, expanding working integrations and refining the underlying interaction model, though the device today is generally understood and marketed with more modest, specific claims about what it can automate rather than the sweeping original LAM pitch.

Are AI wearables a privacy risk in the workplace?

Yes, meaningfully so, and this is a live policy question for any organization where employees might bring ambient-recording devices like Plaud, Bee, or Omi into meetings, client calls, or shared workspaces. These devices are built to capture spoken conversation continuously or on-demand, which raises the same consent and disclosure issues as any recording device, compounded by the fact that captured audio typically gets processed by a third-party AI service in the cloud, not kept purely local. Organizations without a clear policy risk situations where sensitive conversations — HR discussions, client negotiations, legal matters — get recorded and stored on a vendor's servers without everyone in the room understanding that. The sensible response isn't necessarily banning these devices outright, since the productivity case for accurate meeting capture is real, but establishing clear rules: disclosure requirements, restricted contexts (no recording in HR or legal conversations), and vetting of where the vendor actually stores and processes the captured data.

What's the battery life on a typical AI pin/pendant device?

Battery life varies meaningfully by device and use case, and it was one of the recurring practical criticisms of the original "do everything" pins — an all-day wearable that can't reliably last through a workday undermines its own core premise. The available research doesn't give precise battery-life figures across every device in this category, so rather than inventing specific numbers, the honest general guidance is: narrower-purpose devices built around on-demand recording (like Plaud, Bee, or Omi, which activate primarily during active use rather than running continuous always-on AI processing) tend to manage battery life more sustainably than devices attempting continuous ambient AI assistance throughout the day, which is part of why the "always-on do-everything assistant" framing proved harder to deliver reliably at the hardware level than the narrower, more intermittent-use transcription and capture devices that have found commercial traction.

Do AI pin devices work without an internet connection?

Generally, no — most of the devices discussed here, including Plaud, Rabbit R1, Omi, and Bee, depend on cloud connectivity to run the AI processing (transcription, summarization, language understanding) that makes them useful, rather than performing that processing entirely on-device. This is standard across the ambient-AI wearable category broadly, since the language models involved are typically too large and computationally demanding to run efficiently on small, battery-constrained hardware. Practically, this means these devices usually need either their own cellular/data connection or a paired connection through a smartphone via Bluetooth or Wi-Fi to function as advertised, and it's a relevant consideration for anyone planning to use one somewhere with unreliable connectivity — the device may still capture raw audio locally in some cases, but transcription, summarization, and AI features will typically be delayed or unavailable until connectivity is restored.

How does Plaud's NotePin compare to using a smartphone voice-memo app?

The core functional difference is dedicated design versus general-purpose convenience. A smartphone voice-memo app is free, always available, and requires no extra hardware — but it typically just records raw audio, leaving transcription, summarization, and multi-language support to whatever separate app or manual effort you add afterward. Plaud's NotePin is a $159–$189 dedicated device built specifically around accurate transcription across 112 languages with integrated AI summarization, aimed at people for whom meeting and conversation documentation is frequent enough to justify a purpose-built tool. For occasional, casual use, a phone's built-in voice memo app is probably sufficient. For professionals — consultants, salespeople, researchers, journalists — who need reliable, searchable, multi-language transcripts as a regular part of their workflow, the dedicated device's reported million-plus-unit adoption suggests a meaningful number of users have concluded the specialized tool is worth the extra hardware and cost over the free built-in alternative.

What did the original Rabbit R1 launch get criticized for?

The original 2024 Rabbit R1 launch drew heavy criticism for a gap between its marketing promise and its actual functionality. The device was pitched around its "Large Action Model," positioned as capable of learning to operate a broad range of apps and services on the user's behalf, but reviewers — including a widely discussed MKBHD review describing the device as barely functional at launch — found that only a small handful of integrations actually worked reliably at the time of release, far short of the sweeping automation promise in the marketing. This combination of unproven core technology, limited working functionality, and ambitious framing echoed criticisms leveled at Humane's AI Pin around the same period, reinforcing a broader narrative about the category overselling capability relative to what could actually be delivered. Rabbit's subsequent RabbitOS 2 update in September 2025 was a direct response to this criticism, meaningfully narrowing the gap between promise and delivery.

Is there a resale market for discontinued devices like the Humane AI Pin?

Discontinued AI hardware devices like the Humane AI Pin typically see resale value collapse once their cloud backend and support are wound down, because the hardware alone — without the AI processing service it depended on — usually can't perform its core functions. Available research doesn't provide specific resale price data for the Humane AI Pin post-discontinuation, so rather than inventing figures, the general pattern worth understanding is this: devices in this category are heavily service-dependent, meaning their secondhand value is tied far more closely to whether the original company (or an acquirer) keeps the backend running than to the physical hardware's condition. This is a useful general caution for prospective buyers of any early-stage AI hardware product — evaluate not just the device and its current price, but how dependent its usefulness is on an ongoing cloud service, and what happens to your purchase if that company is acquired or shuts down, as happened with both Humane and Limitless.

Which AI wearable has the best data-privacy practices?

Available research doesn't provide a detailed, verified comparison of data-privacy practices across Plaud, Rabbit R1, Omi, and Bee, so rather than ranking them on unconfirmed claims, it's more useful to point to what actually differentiates their privacy posture structurally. Omi's support for bring-your-own-API-key configurations is notable because it gives users more direct control over which AI provider processes their data, rather than routing everything exclusively through one vendor's infrastructure — a meaningful structural difference from more closed, turnkey devices. Beyond that, anyone evaluating these devices on privacy grounds should look directly at each vendor's published privacy policy and data-retention terms rather than relying on general reputation, given how much these devices depend on cloud processing of what is often sensitive spoken conversation content, and given how quickly this product category has shown itself capable of changing hands between companies (Humane to HP, Limitless to Meta) with unclear implications for how previously collected data gets handled.

Will Meta build its own AI pendant now that it owns Limitless's technology?

There's no confirmed public roadmap on this, so any specific answer would be speculation — what is confirmed is that Meta absorbed Limitless's underlying technology into its broader wearable efforts after acquiring the company and shutting down the standalone consumer pendant service in December 2025. Given Meta's substantial existing investment in wearables — its Ray-Ban smart glasses line and reported work on more advanced future devices — it's plausible the ambient-AI and audio-capture technology Limitless built could resurface as a feature within Meta's existing or future wearable products rather than as a separate standalone pendant. But that's an inference based on Meta's general strategic pattern and existing product lines, not a confirmed plan, and readers should treat any specific prediction about a future Meta pendant product as speculative rather than established fact.