From Sci-Fi to Reality: The History of AI Wearables and What Comes Next

The idea of a personal AI assistant you wear on your body has been a staple of science fiction for decades. From Star Trek's communicator badges to Iron Man's JARVIS, the vision was always the same: an intelligent companion that's always available, always listening, always ready to help.

The Early Days (2023-2024)

The first wave of AI wearables were ambitious but flawed. Limited battery life, poor audio quality, and cloud-dependent processing meant early adopters endured significant friction. But they proved the concept: people genuinely wanted an AI assistant that lived in the physical world, not behind a screen.

The Maturation (2025-2026)

The current generation has solved most first-wave problems. Multi-day battery life, on-device processing, and refined AI models have turned wearables from novelty items into daily essentials. The market has also segmented: some devices optimize for meetings, others for health, others for general memory augmentation.

What's Next

The next frontier is proactive intelligence. Instead of waiting for you to ask questions, future wearables will anticipate your needs based on context. Imagine a device that knows you're walking into a meeting with someone you met six months ago and whispers a quick briefing about your last conversation. That's not science fiction anymore — it's engineering.

Lessons from each generation of failed and successful wearables

The 1960s and 70s gave us the calculator watch, the first hint that compute could ride on the wrist. The 1980s brought the Casio Databank and the early experiments by Steve Mann at MIT, who built head-mounted cameras and displays decades before any consumer cared. These prototypes failed commercially but established the form factors and the questions that every subsequent generation has wrestled with: how do you balance utility, social acceptability, and battery life within a few cubic centimeters?

Google Glass in 2013 was the cautionary tale that defined the next decade. The hardware was actually impressive for its era, but the social rollout was a disaster: the glasshole meme, restaurant bans, and privacy backlash killed the consumer product. The lesson was that wearables are uniquely visible and uniquely intrusive, and that public consent matters as much as technical capability. Every smart glass team since, from Snap to Meta to Apple, has been working through the implications of that failure.

Fitbit and the first Apple Watch generations rewrote the playbook. Quiet, ambient, focused on a single domain (health), and worn somewhere people already wore jewelry. By picking a constrained use case and executing it well, both companies built the consumer trust that more ambitious form factors had squandered. The lesson here was that wearable success is iterative: start narrow, earn the right to expand, and let users discover the ecosystem on their own schedule rather than demanding revolution on launch day.

The 2023 to 2024 AI hardware wave, including Humane, Rabbit, and the first pendant startups, made the opposite mistake. They positioned themselves as smartphone replacements before they had earned the right, and the products shipped with rough software, short battery life, and uneven AI quality. Each failure was instructive. Humane proved that an AI device needs a screen for some tasks. Rabbit revealed how dependent these products are on third-party app integrations that companies cannot control. The graveyard taught the survivors what not to do.

Looking forward, three threads seem most likely to define the next five years. Smart glasses with passable displays, building on the Ray-Ban Meta foundation. Ambient pendants and pins focused on memory and capture rather than full assistance. And health-first rings and bands that quietly add AI features without changing the form factor. The companies that win will be the ones that resist the temptation to ship a device that does everything, and instead build something that does one thing so well that wearing it feels obvious within a week.

Why the smartphone slowed wearable progress

For nearly a decade after the iPhone launched in 2007, the smartphone absorbed almost every new sensor and feature that might have lived on the wrist or face. Pedometers, GPS, heart-rate proxies, voice assistants like Siri and Google Now, even barometers all landed inside Apple and Samsung flagships. That gravitational pull starved early wearables of differentiation, which is why Google Glass in 2013 and the first Pebble in 2012 felt like accessories rather than categories. The Apple Watch Series 0 in 2015 only escaped that fate because watchOS 2 finally cut the iPhone tether for core fitness tracking.

AI changed that calculus around 2023, when on-device models became small enough to run useful tasks without round-tripping to the cloud. Qualcomm's Snapdragon W5+ for wearables and Apple's S9 chip with a four-core Neural Engine made local inference plausible inside a watch case. Suddenly the phone was no longer the only sensible home for assistants, transcription, or contextual nudges. Wearables had a reason to exist that did not require the user to pull out their phone, which is the same shift that made Bluetooth earbuds dominant after the iPhone 7 removed the headphone jack in 2016.

The form-factor wars of 2024

Between the Humane AI Pin launch in April 2024 and the Rabbit R1 debut at CES the same year, the industry briefly convinced itself that screenless and phone-replacement devices were the future. Both products collected harsh reviews from Marques Brownlee and The Verge within weeks, with Humane recalling chargers for a fire risk and Rabbit accused of running a wrapped Android app. By the end of 2024 Humane sold its remaining assets to HP for around 116 million dollars, and the lesson was clear. Replacing the phone is harder than augmenting it.

Meta and Ray-Ban took the opposite bet with the second-generation Ray-Ban Meta glasses, which shipped in October 2023 and crossed two million units sold by mid-2025 according to EssilorLuxottica's earnings call. The glasses had no display, leaned on a camera and Meta AI, and looked like normal Wayfarers. That restraint mattered. Fieldy, Friend, and the Bee pendant followed similar logic in 2025, accepting that a useful AI companion did not need a screen if it could listen, remember, and answer through a paired phone.

How regulation is shaping the next wave

The EU AI Act entered into force on August 1, 2024, and its general-purpose model obligations kicked in on August 2, 2025. Wearables that record bystanders, like camera glasses or always-listening pendants, now sit close to the high-risk category if they process biometric data. Meta's Ray-Ban glasses ship with a capture LED for exactly this reason, and Fieldy's published privacy policy commits to on-device processing for raw audio. California's AB 1008, which extended CCPA to neural data in 2024, adds another layer that any US shipping wearable now has to plan around.

Hardware roadmaps are bending to match. Qualcomm's AI Hub published in 2024 now lists wearable-specific models under 1 billion parameters, including a quantized Whisper Tiny that runs at roughly 30 milliseconds per second of audio on a Snapdragon W5 Gen 1. Apple's October 2025 Watch Series 11 documentation confirms that Apple Intelligence transcription stays on-device unless the user explicitly opts into Private Cloud Compute. The pattern is consistent. Regulators want auditable data flows, and silicon vendors are giving device makers the tools to ship them without sacrificing useful AI features.

The battery problem nobody solved

Every generation of AI wearable has hit the same wall. Continuous audio capture and any local inference budget burn through small lithium-polymer cells fast. The Humane AI Pin promised a full day and delivered four to six hours of active use, with a battery booster you had to swap mid-day. The Rabbit R1 fared similarly. Even the Ray-Ban Meta glasses, with a charging case modeled on AirPods Pro, limit recording sessions to about 60 minutes before the temples overheat. This is not a software problem, it is energy density, and Wright's Law on lithium cells has been crawling.

The interesting workarounds are coming from two directions. First, silicon. The Ambiq Apollo5, announced in 2024, claims sub-microamp standby and pushes always-on voice keyword spotting into a fraction of the power budget that an Arm Cortex-M55 needs. Second, design. Fieldy, the Bee pendant, and Limitless Pendant all moved the heavy compute to a paired phone over Bluetooth Low Energy, treating the wearable as a microphone with manners. Until solid-state batteries from QuantumScape or Samsung SDI reach consumer scale, that split-architecture model is the only honest path to all-day AI on the body.

What developers are actually building

Strip away the keynote demos and the working software in 2025 and 2026 falls into a small number of buckets. Memory and recall is the largest, covering Limitless, Fieldy, Bee, and Rewind's pivot toward audio. Translation is the second, with Timekettle X1 and the Ray-Ban Meta live translation feature that shipped in April 2025. Health passive monitoring is the third, anchored by the Oura Ring 4 and Whoop 5.0, both of which added GPT-style natural language summaries last year. Each bucket has a clear job, and the products that picked one are the ones that lasted.

What is missing, and what the next two years will likely deliver, is the connective tissue. Open-source projects like Personal AI's OPI protocol and Apple's App Intents framework, expanded in iOS 18 and iOS 19, let a wearable's memory feed Shortcuts, Notion, or a family calendar without bespoke integrations. The companies that nail this layer will look less like gadget makers and more like operating systems. That is also why Fieldy's roadmap leans hard on calendar and todo integrations, because a memory you cannot act on is just a podcast you do not have time to listen to.

Predictions for 2027 and beyond

Three things look likely by late 2027. First, glasses will eat the smartwatch's lunch for ambient AI. Meta's Orion prototype, demoed at Connect 2024, plus Apple's rumored second-generation Vision product targeting under 1500 dollars, will pull the heads-up display category into the mainstream. Second, audio wearables without screens will not disappear, but they will consolidate. Expect at most three or four credible brands left standing, with Fieldy, Limitless, and a Meta or Apple entrant the most likely survivors. The Humane and Rabbit pattern of overpromising will not repeat, because reviewers like MKBHD have made the cost of overreach very public.

Third, the real unlock will be agentic behavior, not better recall. When a wearable can not only remember that you promised to call your mother on Sunday, but also draft the message, check her calendar through a shared family graph, and surface the right moment to do it, the category moves from neat to necessary. Anthropic's Claude with computer use, released in October 2024, and OpenAI's Operator preview in early 2025 are the early scaffolding. Wearables will be the body those agents borrow. That is the version of the future the next generation of devices is quietly building toward.