×
The best earbuds we’ve tested for 2025

The best earbuds we’ve tested for 2025

Tech-news , , , ,

It’s hard to buy a bad pair of wireless earbuds these days, and with constant discounts and deals wherever you look, now is as good of a time as any to splurge on the pair you’ve been eyeing. The market has come a long way since the early era of true wireless earbuds when we had to deal with mediocre sound quality and unreliable performance, all for the sake of ditching cables. Things are much different now. After several product generations of learned lessons, companies like Sony, Apple, Samsung, and others are releasing their most impressive earbuds to date.

You can get phenomenal noise cancellation and sound quality in the premium tier of earbuds if you’re willing to spend big. But those aren’t always the most important criteria for everyone: maybe you’re looking for the perfect fitness earbuds or for a set that works just as well for Zoom calls as for playing your favorite playlists and podcasts.

Tech companies are increasingly making their earbuds work best with their own products through exclusive features and functionality, so that’s another thing to consider as you shop around. If you want a perk like head-tracking spatial audio, you’ll need to use AirPods with an iPhone, Samsung buds with a Samsung phone, and so on.

A photo of Sony’s WF-1000XM5 earbuds.

$248

Sony’s flagship wireless earbuds offer the company’s best noise cancellation, powerful and lively sound, and a comfortable fit.

A photo of Bose’s QuietComfort Ultra Earbuds.A photo of Bose’s QuietComfort Ultra Earbuds.

$285

With dynamic, rich sound and the best ANC available, the QC Ultra Earbuds are a terrific pick if you want to listen to your music in peace.

What I’m looking for

Whether you’re using earbuds to check out a newly-released album on Spotify, as motivation at the gym, or to tune out distractions around the office, they must deliver on the key metrics of sound quality, noise cancellation, voice call performance, and more. We spend many hours putting our picks through the paces. This includes quiet listening sessions that reveal their full detail, overall fidelity, and dynamics. We also take these buds out into the real world — on rush hour subway commutes, around busy cities, and in crowded cafes — to gauge their connection strength and determine how effectively they can quiet the world around you. We test them on video calls at work and by calling friends and family, who provide blunt feedback on how their mics sound. And importantly, with earbud makers often reserving certain software tricks for their own preferred ecosystem, we see how the earbuds fare when used across multiple platforms.

No matter how they sound, what good are earbuds if you don’t enjoy wearing them? Ears come in all shapes and sizes, so the comfort factor is highly subjective. Most of our picks include several sizes of ear tips; usually they’re silicone, but some companies use foam (or both). Again, which you like better comes down to personal preference. Silicone tips are more common, but foam can expand to fill more of your ear for a better seal. The most important tip I can offer is to try different-sized tips in each ear: you might find that one of your ears needs something bigger or smaller than the other.

Just like with comfort, we all have our own inclinations for what we want from sound quality. Some people are bass-heads who can’t live without a thumping low end. Personally, I care more about dynamics, clarity, and definition from the music I’m listening to. I want vocals to ring through naturally and instruments to be well separated in the sound stage. I look for earbuds that can bring out the subtle details in my favorite songs.

What was once a luxury feature is now the status quo. Many of our top picks include active noise cancellation to help quiet the raucous outside world. Whether you’re trying to concentrate at the office or rest while traveling, noise cancellation can be a huge help. However, some people hate the “plugged up” feeling of earbuds and prefer open-style buds that help maintain full awareness of their surroundings.

Most earbuds these days can reach between five and six hours of continuous battery life with ANC enabled. Some models can exceed that (and over-ear headphones tend to last much longer on a charge), but the convenience of being able to just pop your earbuds back into their case for more juice makes the tradeoff easier to live with.

Even if you don’t intend to use your earbuds for fitness purposes, there’s always a chance of getting caught out in the rain while wearing them. At minimum, you should be looking for earbuds that offer IPX4 water resistance. This safeguards them from splashes of water (and sweat), so they should be cut out for the hazards of daily life.

The best overall wireless earbuds

$248

The Good

  • Smaller, more comfortable design
  • Upgrades to ANC and sound quality
  • Much clearer voice call performance

The Bad

  • Glossy texture can make them harder to remove from case
  • Sony’s bonus features aren’t as seamless as those of its competitors
  • Potential battery longevity concerns

Noise cancellation: Yes / Battery life: 8 hours (ANC on) / Water resistance: IPX4 / Bluetooth codecs: LDAC, AAC, SBC / Multipoint: Yes / Works best with: Android, iOS, Windows, macOS / Ear tips: Foam (XS, S, M, L)

Just like their predecessors, Sony’s WF-1000XM5 earbuds have again topped our recommendations for the overall best noise-canceling earbuds. Thanks to their larger drivers, the audio quality is more detailed and dynamic than before while still retaining the warmth and clarity that made the WF-1000XM4 so enjoyable. Sony has also further improved its ANC and is nearing the same territory as Bose’s QC Earbuds II. These earbuds do a terrific job of eliminating background noise. Battery life can stretch to eight hours of continuous listening with noise cancellation turned on.

Some people found the XM4 to be bulky and uncomfortable, so this time around, Sony has managed to shrink the XM5 by 25 percent. Like before, the company includes foam-style tips that expand in your ears to create an excellent seal — and now there’s a fourth (extra small) set of tips in the box.

A photo of Sony’s WF-1000XM5 earbuds.

The WF-1000XM5 are smaller and more comfortable than their M4 predecessors.
Photo by Chris Welch / The Verge

On Android phones, the 1000XM5 support Sony’s LDAC codec for higher-quality music playback. And they’re capable of dynamic head tracking spatial audio through supported video apps like Netflix. But even if you’re not using LDAC (or have an iPhone), these are up there with the very best audio-focused premium earbuds — including Sennheiser’s Momentum True Wireless 4, only with substantially better noise cancellation. They’re also capable of multipoint Bluetooth connectivity, so you can pair them with two devices at the same time.

It’s worth noting that there were many reports of battery issues with the 1000XM4 over time as those earbuds aged, but Sony claims to have rectified that problem with the new model by tweaking the charging algorithm.

Read our full WF-1000XM5 review.

The best budget earbuds under $100

A photo of new earbuds from Nothing.A photo of new earbuds from Nothing.

$79

The Good

  • Standout color and design
  • Very compact carrying case
  • Great performer for the price

The Bad

  • Limited retail availability
  • Plastic case can show scuffs

Noise cancellation: Yes / Battery life: 5.5 hours (ANC on) / Water resistance: IP54 / Bluetooth codecs: LDAC, AAC, SBC / Multipoint: No / Works best with: Android, iOS, Windows, macOS / Ear tips: Silicone (S, M, L)

When looking at Nothing’s Ear (a) earbuds, it’s easy to focus all your attention on the catchy yellow color or their fun, super compact carrying case. But at their $99.99 price, the company’s entry-level earbuds provide a ton of features, including active noise cancellation, a low-latency mode for gaming, multipoint Bluetooth pairing, and a decent IP54 dust and water resistance rating.

A photo of new earbuds from Nothing.

The Ear (a) are Nothing’s first earbuds that aren’t only offered black or white.
Photo by Chris Welch / The Verge

And the sound quality you get outperforms my usual expectations for this price point. They’re not quite on par with Nothing’s higher-end Ears, but you do get expressive, very enjoyable audio — and you can customize the EQ further in Nothing’s mobile app.

Read my full Nothing Ear (a) impressions.

The best noise-canceling earbuds

$285

The Good

  • Most powerful ANC you’ll find in earbuds
  • Lively, rich sound
  • Stabilizer fins are easier to apply

The Bad

  • Immersive Audio drains battery
  • Still no multipoint in 2023
  • Wireless charging requires $49 case

Noise cancellation: Yes / Battery life: 6 hours (ANC on) / Water resistance: IPX4 / Bluetooth codecs: AAC, SBC, aptX Adaptive / Multipoint: Yes / Works best with: Android, iOS, Windows, macOS / Ear tips: Silicone (S, M, L)

Bose’s QuietComfort Ultra Earbuds are the most powerful noise-canceling earbuds on the market. Of all our top picks, they do the best job at offering peace and quiet from the cacophony of daily life and putting you in a private bubble with your music playlists or favorite podcasts.

A photo of Bose’s QuietComfort Ultra Earbuds.

Bose’s QuietComfort Ultra Earbuds offer the best noise cancellation of any pair of wireless buds.
Photo by Chris Welch / The Verge

The QC Ultra Earbuds are extremely similar to 2022’s QC Earbuds II in terms of design, fit, and sound. Audio quality is rich and dynamic, delivering more bass thump than, say, Apple’s AirPods Pro. And by switching to a two-piece ear tip / stabilizer fitting process, Bose’s latest flagship buds should fit a wider range of ears more comfortably. The Ultra also introduce a new Immersive Audio mode, which is Bose’s take on spatial audio.

Read our full QuietComfort Ultra Earbuds review.

The best earbuds for sports and working out

$249

The Good

  • Huzzah! Wireless charging!
  • Adds ANC, transparency modes, spatial audio
  • Slimmer ear hook, smaller case
  • Adds heart rate
  • Great sound
  • No price hike

The Bad

  • Case is still a chonker
  • Heart rate is not that useful for iOS users

Noise cancellation: Yes / Battery life: 8 hours (ANC on) / Water resistance: IPX4 / Bluetooth codecs: AAC, SBC / Multipoint: No / Works best with: Android, iOS, Windows, macOS / Ear tips: Silicone (XS, S, M, L, XL)

After a long wait, Beats finally introduced the Powerbeats Pro 2 in early 2025. They carry forward the signature ear hook design of past Powerbeats earbuds, but improve upon the original Powerbeats Pro in several key ways. Beats added active noise cancellation and a natural-sounding transparency mode to the Pro 2; the first-generation pair lacked both. The earbuds themselves are lighter, too, and the charging case is noticeably smaller.

The Powerbeats Pro 2 also feature heart rate monitoring thanks to built-in sensors on each earbud. You might already have a fitness tracker, chest strap, or smartwatch for this purpose, but if not, you can use these with compatible apps like Nike Run Club, Runna, Ladder, Slopes, Open, Peloton, and YaoYao. The downside is that they can’t simultaneously play music and broadcast heart rate to gym equipment when paired to an iOS device, rendering the Pro 2 a better heart rate monitoring alternative for Android users than iPhone users.

A photo of a person wearing the Powerbeats Pro 2 earbuds.

The Powerbeats Pro 2 have an ear-hook design for maximum stability.
Photo by Amelia Holowaty Krales / The Verge

The Powerbeats Pro 2 remain rated IPX4 for water resistance. While that’s not as robust as some other fitness-focused buds, it should be sufficient to survive sweat and outdoor runs in various conditions. The earbuds still offer easy-to-use physical buttons and a dedicated volume rocker, so it’s easy to control them in the middle of a workout.

If you’re an iPhone owner, the Powerbeats Pro 2 earbuds include Apple-only features like dynamic head tracking for spatial audio, auto device switching, Find My integration, and more. For Android users, Beats has an app that packs in a smaller subset of these functions.

Read our full Powerbeats Pro 2 review.

The best earbuds for iPhone owners

Apple’s second-generation AirPods Pro photographed on a reflective black surface.Apple’s second-generation AirPods Pro photographed on a reflective black surface.

$199

The Good

  • Much better noise cancellation
  • Improved sound quality and clarity
  • Unrivaled hearing health features

The Bad

  • Unchanged design
  • Sound is better but not top tier

Noise cancellation: Yes / Battery life: 6 hours (ANC on) / Water resistance: IPX4 / Bluetooth codecs: AAC, SBC / Multipoint: No / Works best with: iOS, Windows, macOS / Ear tips: Silicone (XS, S, M, L)

For those invested in Apple’s ecosystem, the second-gen AirPods Pro bring an appealing mix of top-notch noise cancellation and refined sound compared to the original pair. As always, they also include numerous features (spatial audio, audio sharing, automatic syncing across iCloud devices, etc.) that make them an ideal companion to the company’s other devices. They’re also the only AirPods that come with different sizes of ear tips for a customizable fit — including an extra small option.

Apple’s second-generation AirPods Pro pictured in a side profile photo of a woman’s head.

The second-gen AirPods Pro remain the best choice for people who own Apple devices.
Photo by Chris Welch / The Verge

The AirPods Pro have a compact case that supports wireless charging. Setting them up is as simple as holding the open case near your iPhone. The earbuds are lightweight and fit snugly in most ears. Their force sensor pinch controls are easy to get down, and with the second-gen model, Apple added swipe-based volume controls. The AirPods Pro have traditionally ranked among the best wireless earbuds when it comes to voice call quality, and the second-gen earbuds don’t disappoint. Their ANC is improved over the original model and is now right up there with Sony, and their transparency mode has gained the ability to quickly damper sudden, loud noises that might happen around you.

Where the AirPods Pro stand out is with their Apple-only features like head tracking for spatial audio, automatic device switching, extensive Find My support, and so on. The charging case now features a built-in speaker and a U1 chip for pinpoint location tracking, and the latest AirPods Pro have better sound and noise canceling capabilities than the original model. Plus, they retain their deep integration with Apple’s hardware ecosystem.

In the fall, Apple also rolled out an impressive set of hearing health features for the second-gen AirPods Pro. They’re now able to serve as clinical-grade hearing aids for people with mild to moderate hearing loss and protect your hearing in loud environments; you can even take a hearing test to assess your own level of hearing loss.

Read our original AirPods Pro (second-gen) review.

The best earbuds for Samsung Galaxy phone owners

A hands-on photo of Samsung’s Galaxy Buds 3 Pro earbuds.A hands-on photo of Samsung’s Galaxy Buds 3 Pro earbuds.

$190

The Good

  • Terrific sound quality
  • Crisp, natural transparency / passthrough mode
  • Voice commands are simple and so useful
  • Good voice calls

The Bad

  • Uninspired design
  • Midtier noise cancellation performance
  • No true multipoint

Noise cancellation: Yes / Battery life: 6 hours (ANC on) / Water resistance: IP57 / Bluetooth codecs: Samsung seamless, AAC, SBC / Multipoint: No / Works best with: Android, Windows, macOS / Ear tips: Silicone (S, M, L)

Samsung’s Galaxy Buds 3 Pro ditch the company’s more discreet designs of the past for a stemmed style that’s similar to AirPods and any number of other wireless earbuds. Boring? Maybe, but thanks to that change, I’ve found them to be more comfortable and provide better stability than the Buds 2 Pro. The sound quality is even better than before — and it was already excellent. These are right up there with Sennheiser and Technics from an audio fidelity standpoint.

A hands-on product photo of Samsung’s Galaxy Buds 3 / Galaxy Buds 3 Pro.

The Buds 3 Pro are an ideal match for Samsung’s latest phones.
Photo by Chris Welch / The Verge

The Buds 3 Pro are also a standout performer when it comes to call quality. And my favorite thing about these earbuds are the hands-free voice commands that don’t require you to say any specific wake phrase beforehand. You can just say “volume up” and / or “next song,” and it happens.

But they’re not perfect: the Buds 3 Pro lack multipoint connectivity, so they can’t connect to two devices at the same time. Just like Apple, Samsung prefers to automatically hop between products within its own walled garden ecosystem. Some people might consider this a dealbreaker, but I’ve still enjoyed the earbuds plenty without it.

Read our full Galaxy Buds 3 Pro review.

The best earbuds for Google Pixel phone owners

A hands-on photo of Google’s Pixel Buds Pro 2 earbuds.A hands-on photo of Google’s Pixel Buds Pro 2 earbuds.

$189

The Good

  • Smaller and lighter design is supremely comfortable
  • More powerful noise cancellation
  • Crystal-clear transparency mode
  • Very pleasant and detailed sound quality

The Bad

  • ANC can’t quite match Bose
  • Google still refuses to add higher-quality Bluetooth codecs, limiting overall fidelity
  • Gemini Live feels more feeble than futuristic

Noise cancellation: Yes / Battery life: 8 hours (ANC on) / Water resistance: IP54 / Bluetooth codecs: AAC, SBC / Multipoint: Yes / Works best with: Android, Windows, macOS / Ear tips: Silicone (XS, S, M, L)

With the Pixel Buds Pro 2, Google has addressed most of the downsides of the original pair. They’re far smaller and lighter than the chunky first-gen Pixel Buds Pro. And there’s a small nub on the buds to help lock them in place even during intensive activities like running or exercise.

Google’s active noise cancellation still isn’t on the same tier as Bose, but the transparency mode on the Buds Pro 2 is among the very best you’ll find. It offers a crisp, natural passthrough for the outside world when needed. Multipoint support is included, and Google’s latest earbuds eke out impressive battery life that surpasses most competitors.

A hands-on photo of Google’s Pixel Buds Pro 2 earbuds.

Google really downsized the Pixel Buds Pro 2 (bottom) compared to the original pair (top).
Photo by Chris Welch / The Verge

Got a Pixel phone? With the Pixel Buds Pro 2, you can have lengthy conversations with Google’s Gemini AI while your phone stays in your pocket. Features like head-tracking spatial audio and Clear Calling are also exclusive to those within the Pixel ecosystem. If that’s you, the Pixel Buds Pro 2 are a big step up from the previous buds in a much smaller package.

Read our full Pixel Buds Pro 2 review.

The best earbuds if you own both Apple and Android devices

A photo of Beats’ translucent Studio Buds Plus earbuds.A photo of Beats’ translucent Studio Buds Plus earbuds.

$170

The Good

  • Better sound, ANC, and battery life than Studio Buds
  • Standout translucent case option
  • Improved voice call performance

The Bad

  • ANC and transparency fall short of second-gen AirPods Pro
  • Doesn’t offer all Apple ecosystem tricks
  • No wireless charging or in-ear detection

Noise cancellation: Yes / Battery life: 6 hours (ANC on) / Water resistance: IPX4 / Bluetooth codecs: AAC, SBC / Multipoint: No / Works best with: Android, iOS, Windows, macOS / Ear tips: Silicone (XS, S, M, L)

Beats’ Studio Buds Plus are an upgraded, better-performing revision of the Studio Buds released in 2021. This time, the company has made the noise cancellation more powerful while also improving the quality of the transparency mode and extending battery life. If you’re going for style points, it doesn’t get much cooler than the translucent color option. These earbuds are very tiny and rank up with the best in terms of overall comfort.

A photo of Beats’ translucent Studio Buds Plus earbuds.

The translucent color option has a very ‘90s vibe.
Photo by Chris Welch / The Verge

What’s interesting about the Studio Buds Plus is that they use a proprietary chip that allows them to support native software features on both iOS and Android. Plenty of earbuds offer companion apps on each platform, but in the case of the Beats, they work with both Apple’s Find My and Google’s Find My Device. You also get simple one-tap pairing, no matter which operating system you’re using.

The drawback to this dual-ecosystem life is that Apple customers don’t get all the usual bells and whistles that come with AirPods or even the Beats Powerbeats Pro 2; you lose out on audio sharing, head tracking spatial audio, and other tricks. Wireless charging and an ear detection sensor (for auto-pause when a bud is removed) are also MIA, which stings a bit considering the price.

Read our full Beats Studio Buds Plus review.

The best earbuds for sound quality

A photo of Sennheiser’s Momentum True Wireless 4 earbuds in copper.A photo of Sennheiser’s Momentum True Wireless 4 earbuds in copper.

$231

The Good

  • Fantastic sound quality
  • Improved day-to-day reliability
  • Auracast-enabled

The Bad

  • Too pricey for some
  • Noise cancellation does the job, but isn’t anything special
  • Long-term durability remains to be seen

Noise cancellation: Yes / Battery life: 7.5 hours (ANC on) / Water resistance: IP54 / Bluetooth codecs: AptX Adaptive, AptX, AAC, SBC / Multipoint: Yes / Works best with: Android, iOS, Windows, macOS / Ear tips: Silicone (XS, S, M, L)

Sennheiser’s Momentum True Wireless earbuds have always sounded fantastic; that’s never been the issue. But previous models in the series have been tainted a bit by bugs, unsteady performance, and battery reliability problems. The fourth-gen pair finally gets that part right while continuing to offer tremendous, detailed sound quality that stands above nearly all competitors. They’re a joy to listen to. The active noise cancellation can’t compete with the likes of Apple, Sony, and Bose, but if you care more about lush, intricate sound, the MTW4 won’t disappoint in the slightest.

A photo of Sennheiser’s Momentum True Wireless 4 earbuds in copper.

That copper finish is pretty nice.
Photo by Chris Welch / The Verge

These earbuds come with optional wing tips to help keep them securely locked in your ears. That could prove helpful for running or when hitting the gym. Sennheiser allows for plenty of EQ customization using its mobile app, and the Momentum True Wireless 4 support a range of Bluetooth codecs, including AAC, SBC, AptX, and AptX Adaptive. Sennheiser has also promised that they’re LE Audio and Auracast-ready for when those features become more widespread.

Read our full Momentum True Wireless 4 review.

The best earbuds for voice calls on iPhone

A hands-on photo of Apple’s AirPods 4 wireless earbuds.A hands-on photo of Apple’s AirPods 4 wireless earbuds.

$119

The Good

  • Upgraded sound quality
  • Voice isolation can noticeably improve call clarity
  • Deep Apple ecosystem integration

The Bad

  • No wireless charging
  • No built-in speaker on the case
  • No proper multipoint

Noise cancellation: Optional / Battery life: 5 hours / Water resistance: IP54 / Bluetooth codecs: AAC, SBC / Multipoint: No / Works best with: iOS, Windows, macOS / Ear tips: None

Apple’s AirPods 4 come in two different versions, with a higher-priced option that includes active noise cancellation and wireless charging. But sound quality is identical on both sets, and equally as important in this case is that the same can be said of voice call clarity.

On the AirPods 4, there’s a new “voice isolation” feature — enabled by default — that uses machine learning both on the earbuds and your iPhone to eliminate distracting background noise on your end so that you come through clearly to whoever you’re speaking with.

A hands-on photo of Apple’s AirPods 4 wireless earbuds.

Photo by Chris Welch / The Verge

Overall call quality is excellent with the AirPods 4. And beyond that, compared to the third-gen model, you can expect upgraded audio quality (with richer bass), new ways of interacting with Siri through head movements, and yes… there’s a USB-C connector on the included charging case, as opposed to a Lightning port.

The AirPods still have a one-size-fits-most hard plastic form factor, so I recommend trying them before committing to a purchase. However, Apple says it used more than 50 million data points to further refine the contours and overall stability of its latest earbuds. So, if previous models worked well for you, these seem like a safe bet. The only question is whether you feel like spending a little more to gain ANC, a transparency mode, wireless charging, and a built-in speaker for Find My location tracking.

Read our full AirPods 4 review.

The best earbuds for voice calls on Android

A hands-on photo of Samsung’s Galaxy Buds 3 Pro earbuds.A hands-on photo of Samsung’s Galaxy Buds 3 Pro earbuds.

$190

The Good

  • Terrific sound quality
  • Crisp, natural transparency / passthrough mode
  • Voice commands are simple and so useful
  • Good voice calls

The Bad

  • Uninspired design
  • Midtier noise cancellation performance
  • No true multipoint

Noise cancellation: Yes / Battery life: 6 hours (ANC on) / Water resistance: IP57 / Bluetooth codecs: Samsung seamless, AAC, SBC / Multipoint: No / Works best with: Android, Windows, macOS / Ear tips: Silicone (S, M, L)

Sony’s unconventional LinkBuds were our longtime pick for the clearest voice calls on Android, but we haven’t had the chance to review their successors. For now, the aforementioned Galaxy Buds 3 Pro are an easy-to-recommend choice. The built-in “voice pickup unit” detects the movement of your mouth to help isolate your voice from background noise and keep conversations clear.

A hands-on product photo of Samsung’s Galaxy Buds 3 / Galaxy Buds 3 Pro.

The advantage of stemmed earbuds is that they position the mics closer to your mouth.

The Buds 3 Pro have proven dependable for me during video chats on Google Meet, and people have been able to hear me without issue, even when I’m answering a quick call on the street.

Read our full Galaxy Buds 3 Pro review.

Photography by Chris Welch / The Verge except where specified

Update, June 5th: Updated to reflect current pricing and availability, along with several new links.

Source link
#earbuds #weve #tested

Tag

Related Posts

Federal prosecutors charged a Google employee with fraud after he allegedly made $1.2 million on Polymarket bets related to Search-related trends in 2025, as reported earlier by ABC News. In their now-unsealed complaint, prosecutors allege that Michele Spagnuolo “knew the outcome of these wagers before the trading public did because he had accessed Google’s confidential, commercially valuable internal data.” Spagnuolo was arrested in New York on Wednesday but released on a $2.25 million bond, ABC News reports. He is charged with commodities fraud, wire fraud, and money laundering.

Spagnuolo made bets on Polymarket under the username AlphaRacoon, with his successful search-related wagers catching the attention of outlets like Forbes and users on social media last December. In one instance, Spagnuolo correctly guessed that a singer named D4vd would “be the #1 searched person on Google” in 2025, despite the “near-zero probability” assigned by Polymarket, according to the complaint.

At the same time, Spagnuolo allegedly bet that Pope Leo XIV and Kendrick Lamar would not appear on Google’s “Year in Search 2025” lists, which are difficult to predict because of how they’re calculated. Google says it ranked last year’s terms based on which ones saw the “highest increase in traffic” — not the highest number of searches — between January 1st, 2025 and November 25th, 2025. “By measuring the spike in interest rather than the total number of searches, we can identify the trends that were unique to 2025.”

“Once he won, Spagnuolo then took deliberate steps to conceal his unlawful use of nonpublic information by attempting to obscure the source and ownership of his unlawful proceeds,” the complaint says. Last month, federal prosecutors charged US Army soldier Gannon Ken Van Dyke with fraud for allegedly making a $400,000 Polymarket bet on the capture of Venezuelan President Nicolás Maduro.

In a statement on X, Polymarket called itself “the enforcement leader,” saying its “market integrity infrastructure” flagged Spagnuolo’s activity. “Blockchain trading is transparent, traceable, and bad actors leave footprints,” the company writes, without noting whether the people putting their money down know that.

”We’re working with law enforcement on their investigation,” Google spokesperson Jaclyn Vazquez says in a statement to The Verge. “The employee accessed our marketing material using a tool available to all employees, but using such confidential information to place bets is a serious breach of our policies. We’ve placed the employee on leave and will take the appropriate action.”

#Google #employee #allegedly #information #win #million #PolymarketBusiness,Google,Policy,Tech">A Google employee allegedly used inside information to win .2 million on Polymarket Federal prosecutors charged a Google employee with fraud after he allegedly made .2 million on Polymarket bets related to Search-related trends in 2025, as reported earlier by ABC News. In their now-unsealed complaint, prosecutors allege that Michele Spagnuolo “knew the outcome of these wagers before the trading public did because he had accessed Google’s confidential, commercially valuable internal data.” Spagnuolo was arrested in New York on Wednesday but released on a .25 million bond, ABC News reports. He is charged with commodities fraud, wire fraud, and money laundering.Spagnuolo made bets on Polymarket under the username AlphaRacoon, with his successful search-related wagers catching the attention of outlets like Forbes and users on social media last December. In one instance, Spagnuolo correctly guessed that a singer named D4vd would “be the #1 searched person on Google” in 2025, despite the “near-zero probability” assigned by Polymarket, according to the complaint.At the same time, Spagnuolo allegedly bet that Pope Leo XIV and Kendrick Lamar would not appear on Google’s “Year in Search 2025” lists, which are difficult to predict because of how they’re calculated. Google says it ranked last year’s terms based on which ones saw the “highest increase in traffic” — not the highest number of searches — between January 1st, 2025 and November 25th, 2025. “By measuring the spike in interest rather than the total number of searches, we can identify the trends that were unique to 2025.”“Once he won, Spagnuolo then took deliberate steps to conceal his unlawful use of nonpublic information by attempting to obscure the source and ownership of his unlawful proceeds,” the complaint says. Last month, federal prosecutors charged US Army soldier Gannon Ken Van Dyke with fraud for allegedly making a 0,000 Polymarket bet on the capture of Venezuelan President Nicolás Maduro.In a statement on X, Polymarket called itself “the enforcement leader,” saying its “market integrity infrastructure” flagged Spagnuolo’s activity. “Blockchain trading is transparent, traceable, and bad actors leave footprints,” the company writes, without noting whether the people putting their money down know that.”We’re working with law enforcement on their investigation,” Google spokesperson Jaclyn Vazquez says in a statement to The Verge. “The employee accessed our marketing material using a tool available to all employees, but using such confidential information to place bets is a serious breach of our policies. We’ve placed the employee on leave and will take the appropriate action.”#Google #employee #allegedly #information #win #million #PolymarketBusiness,Google,Policy,Tech

reported earlier by ABC News. In their now-unsealed complaint, prosecutors allege that Michele Spagnuolo “knew the outcome of these wagers before the trading public did because he had accessed Google’s confidential, commercially valuable internal data.” Spagnuolo was arrested in New York on Wednesday but released on a $2.25 million bond, ABC News reports. He is charged with commodities fraud, wire fraud, and money laundering.

Spagnuolo made bets on Polymarket under the username AlphaRacoon, with his successful search-related wagers catching the attention of outlets like Forbes and users on social media last December. In one instance, Spagnuolo correctly guessed that a singer named D4vd would “be the #1 searched person on Google” in 2025, despite the “near-zero probability” assigned by Polymarket, according to the complaint.

At the same time, Spagnuolo allegedly bet that Pope Leo XIV and Kendrick Lamar would not appear on Google’s “Year in Search 2025” lists, which are difficult to predict because of how they’re calculated. Google says it ranked last year’s terms based on which ones saw the “highest increase in traffic” — not the highest number of searches — between January 1st, 2025 and November 25th, 2025. “By measuring the spike in interest rather than the total number of searches, we can identify the trends that were unique to 2025.”

“Once he won, Spagnuolo then took deliberate steps to conceal his unlawful use of nonpublic information by attempting to obscure the source and ownership of his unlawful proceeds,” the complaint says. Last month, federal prosecutors charged US Army soldier Gannon Ken Van Dyke with fraud for allegedly making a $400,000 Polymarket bet on the capture of Venezuelan President Nicolás Maduro.

In a statement on X, Polymarket called itself “the enforcement leader,” saying its “market integrity infrastructure” flagged Spagnuolo’s activity. “Blockchain trading is transparent, traceable, and bad actors leave footprints,” the company writes, without noting whether the people putting their money down know that.

”We’re working with law enforcement on their investigation,” Google spokesperson Jaclyn Vazquez says in a statement to The Verge. “The employee accessed our marketing material using a tool available to all employees, but using such confidential information to place bets is a serious breach of our policies. We’ve placed the employee on leave and will take the appropriate action.”

#Google #employee #allegedly #information #win #million #PolymarketBusiness,Google,Policy,Tech">A Google employee allegedly used inside information to win $1.2 million on Polymarket 

Federal prosecutors charged a Google employee with fraud after he allegedly made $1.2 million on Polymarket bets related to Search-related trends in 2025, as reported earlier by ABC News. In their now-unsealed complaint, prosecutors allege that Michele Spagnuolo “knew the outcome of these wagers before the trading public did because he had accessed Google’s confidential, commercially valuable internal data.” Spagnuolo was arrested in New York on Wednesday but released on a $2.25 million bond, ABC News reports. He is charged with commodities fraud, wire fraud, and money laundering.

Spagnuolo made bets on Polymarket under the username AlphaRacoon, with his successful search-related wagers catching the attention of outlets like Forbes and users on social media last December. In one instance, Spagnuolo correctly guessed that a singer named D4vd would “be the #1 searched person on Google” in 2025, despite the “near-zero probability” assigned by Polymarket, according to the complaint.

At the same time, Spagnuolo allegedly bet that Pope Leo XIV and Kendrick Lamar would not appear on Google’s “Year in Search 2025” lists, which are difficult to predict because of how they’re calculated. Google says it ranked last year’s terms based on which ones saw the “highest increase in traffic” — not the highest number of searches — between January 1st, 2025 and November 25th, 2025. “By measuring the spike in interest rather than the total number of searches, we can identify the trends that were unique to 2025.”

“Once he won, Spagnuolo then took deliberate steps to conceal his unlawful use of nonpublic information by attempting to obscure the source and ownership of his unlawful proceeds,” the complaint says. Last month, federal prosecutors charged US Army soldier Gannon Ken Van Dyke with fraud for allegedly making a $400,000 Polymarket bet on the capture of Venezuelan President Nicolás Maduro.

In a statement on X, Polymarket called itself “the enforcement leader,” saying its “market integrity infrastructure” flagged Spagnuolo’s activity. “Blockchain trading is transparent, traceable, and bad actors leave footprints,” the company writes, without noting whether the people putting their money down know that.

”We’re working with law enforcement on their investigation,” Google spokesperson Jaclyn Vazquez says in a statement to The Verge. “The employee accessed our marketing material using a tool available to all employees, but using such confidential information to place bets is a serious breach of our policies. We’ve placed the employee on leave and will take the appropriate action.”

#Google #employee #allegedly #information #win #million #PolymarketBusiness,Google,Policy,Tech
Rivian has finally revealed that the first customers of the company’s new R2 SUV will get their vehicles on June 9.

The automaker has spent the last few months ramping up its efforts to release the R2, which is more affordable and aimed at a larger market than its current R1 lineup. The new SUV will initially be available in a trim that starts just under $60,000, though Rivian has announced plans to release a “standard” version that starts at $48,490 in 2027.

The company has teased an even more affordable version “starting around $45,000” late next year — a price tag Rivian has promoted since the R2 reveal in 2024.

Rivian has high expectations for the R2. Founder and CEO RJ Scaringe has said it is “maybe the most important thing we’ve launched to date.” The company is betting on an extremely fast ramp-up, with as many as 25,000 vehicles delivered by the end of this year. Ultimately, Rivian hopes the R2 and its hatchback sibling, the R3, will help the company turn a profit for the first time since its founding in 2009.

#Rivian #deliver #SUVs #June #TechCrunchelectric vehicles,EVs,Rivian">Rivian will deliver the first R2 SUVs on June 9 | TechCrunch Rivian has finally revealed that the first customers of the company’s new R2 SUV will get their vehicles on June 9. The automaker has spent the last few months ramping up its efforts to release the R2, which is more affordable and aimed at a larger market than its current R1 lineup. The new SUV will initially be available in a trim that starts just under ,000, though Rivian has announced plans to release a “standard” version that starts at ,490 in 2027. The company has teased an even more affordable version “starting around ,000” late next year — a price tag Rivian has promoted since the R2 reveal in 2024. Rivian has high expectations for the R2. Founder and CEO RJ Scaringe has said it is “maybe the most important thing we’ve launched to date.” The company is betting on an extremely fast ramp-up, with as many as 25,000 vehicles delivered by the end of this year. Ultimately, Rivian hopes the R2 and its hatchback sibling, the R3, will help the company turn a profit for the first time since its founding in 2009. #Rivian #deliver #SUVs #June #TechCrunchelectric vehicles,EVs,Rivian

revealed that the first customers of the company’s new R2 SUV will get their vehicles on June 9.

The automaker has spent the last few months ramping up its efforts to release the R2, which is more affordable and aimed at a larger market than its current R1 lineup. The new SUV will initially be available in a trim that starts just under $60,000, though Rivian has announced plans to release a “standard” version that starts at $48,490 in 2027.

The company has teased an even more affordable version “starting around $45,000” late next year — a price tag Rivian has promoted since the R2 reveal in 2024.

Rivian has high expectations for the R2. Founder and CEO RJ Scaringe has said it is “maybe the most important thing we’ve launched to date.” The company is betting on an extremely fast ramp-up, with as many as 25,000 vehicles delivered by the end of this year. Ultimately, Rivian hopes the R2 and its hatchback sibling, the R3, will help the company turn a profit for the first time since its founding in 2009.

#Rivian #deliver #SUVs #June #TechCrunchelectric vehicles,EVs,Rivian">Rivian will deliver the first R2 SUVs on June 9 | TechCrunch

Rivian has finally revealed that the first customers of the company’s new R2 SUV will get their vehicles on June 9.

The automaker has spent the last few months ramping up its efforts to release the R2, which is more affordable and aimed at a larger market than its current R1 lineup. The new SUV will initially be available in a trim that starts just under $60,000, though Rivian has announced plans to release a “standard” version that starts at $48,490 in 2027.

The company has teased an even more affordable version “starting around $45,000” late next year — a price tag Rivian has promoted since the R2 reveal in 2024.

Rivian has high expectations for the R2. Founder and CEO RJ Scaringe has said it is “maybe the most important thing we’ve launched to date.” The company is betting on an extremely fast ramp-up, with as many as 25,000 vehicles delivered by the end of this year. Ultimately, Rivian hopes the R2 and its hatchback sibling, the R3, will help the company turn a profit for the first time since its founding in 2009.

#Rivian #deliver #SUVs #June #TechCrunchelectric vehicles,EVs,Rivian
Smart starts with the circuit board, not the cloud

Most coverage of smart devices jumps straight to AI features and voice assistants. But the foundation is physical. A device is a printed circuit board, a microcontroller, a fistful of sensors, a radio, and a battery, all crammed into a shell that has to survive being dropped, sat on, and left in a hot car.

This is where hardware development does its quiet, unglamorous work. Engineers pick a microcontroller based on how much computing the device needs versus how little power it can afford to burn. They route signal traces on the board so a Wi-Fi radio doesn’t drown out a delicate sensor reading. They run the whole thing through thermal testing, drop testing, and certification for FCC and CE marks before it can legally ship.

Get this layer wrong, and no amount of clever software saves you. A poorly designed board produces flaky sensor data. Bad antenna placement means the device drops off your network the moment you walk to the next room. These aren’t software bugs. You can’t patch your way out of a physics problem.

The companies building good hardware treat the proof-of-concept stage as a real checkpoint. They wire up development boards and modular parts to test the core idea cheaply, before committing to a custom design that costs real money to manufacture. It’s the boring discipline that separates products from expensive paperweights.

Firmware is where the device actually thinks

Sitting on top of the hardware is firmware. This is the low-level code that tells the chip what to do, when to wake up, how to read a sensor, and when to phone home. People mix up firmware and software all the time, so here’s the clean split. Software runs on your phone or in the cloud and handles the screens you tap. Firmware lives inside the device and controls the hardware directly.

Firmware is genuinely hard to write well. The constraints are brutal. A typical IoT microcontroller has a tiny amount of memory, often measured in kilobytes, and it might run on a coin cell that needs to last a year. Every line of code competes for space and power.

Then there’s timing. A lot of devices need deterministic, real-time behavior, meaning a sensor reading has to be processed within a fixed window or the whole thing falls apart. A heart monitor that processes a beat “eventually” is useless. The firmware has to guarantee it happens now.

If you want the deep version of how this gets built in practice, Yalantis published a solid breakdown of firmware development for embedded IoT devices that covers architecture, power management, and the over-the-air update workflows that keep a device current after it ships. The OTA piece matters more than it sounds. A device that can’t safely update its own firmware is frozen in time the day it leaves the factory.

Connectivity is a series of trade-offs

Your smart device has to talk to something. Your phone, your router, a cloud server, or all three. Choosing how it talks is one of the most consequential engineering calls in the whole project, and there’s no single right answer.

Bluetooth Low Energy sips power and works great for a wearable talking to your phone, but its range is short and it can’t reach the internet on its own. Wi-Fi reaches everything but drains batteries fast. LoRaWAN travels for miles on almost no power, which is perfect for a soil sensor in a field, but it carries tiny amounts of data slowly. Cellular options like NB-IoT and LTE-M let a device work anywhere there’s a signal, with the catch of ongoing data costs and bigger power draw.

Engineers usually mix these. A fitness band might use BLE to sync with your phone, and your phone carries the data the rest of the way. An industrial sensor in a remote location might use LoRaWAN to a gateway, which then forwards everything over cellular. The “right” combination depends entirely on power budget, data volume, range, and cost, which is exactly why this decision gets made early and gets revisited often.

Sensors and the messy job of trusting them

A smart device is only as good as the data it collects. And raw sensor data is messy.

Take a simple temperature reading. The sensor drifts over time. It gets warmed by the heat of the chip sitting next to it. It returns noisy values that jitter up and down even when nothing changes. Firmware has to calibrate, filter, and sanity-check all of it before the device acts on a single number.

This gets serious fast in regulated fields. A continuous glucose monitor or a medical wearable can’t ship a reading that’s “close enough.” The sensor design, the calibration, and the firmware that validates the data all have to meet standards that consumer gadgets never face. The engineering bar is much higher, and the cost of getting it wrong is measured in patient safety, not customer reviews.

For everyday devices the stakes are lower, but the principle holds. Good devices spend a lot of hidden effort turning unreliable physical signals into numbers you can actually trust.

Where the AI hype meets the silicon

Here’s the part that has changed most recently. A growing share of smart devices now run machine learning models directly on the chip instead of sending everything to the cloud. This is edge computing, and it’s reshaping how devices get built.

The appeal is obvious. Processing data on the device means lower latency, since you’re not waiting on a round trip to a server. It means better privacy, because your data never leaves your hand. And it means the device keeps working when your internet goes down.

The catch is that running a model on a chip with kilobytes of memory is an engineering puzzle. Models have to be shrunk, quantized, and optimized until they fit in the space available without melting the battery. The face-recognition that runs locally on a modern doorbell is a heavily compressed version of what would run on a server. Squeezing it down to fit is real, specialized work, and it’s increasingly where the competitive difference between two similar gadgets actually lives.

Security can’t be the last step

For years, connected devices treated security as an afterthought. Ship the product, patch problems later. That approach has aged badly.

Outdated firmware is now one of the most common ways attackers break into IoT systems. Research from the security firm ONEKEY found that vulnerable firmware accounts for a large majority of successful attacks on connected devices. Once an attacker is inside one poorly secured gadget on your network, they have a foothold to reach everything else.

Building security in from the start means encrypting data both when it’s stored on the device and when it travels to the cloud. It means signing firmware updates so a device only accepts legitimate code, not something an attacker swapped in. And it means designing for recovery, so a compromised device can be safely reset and restored rather than turned into a permanent liability sitting on your shelf.

This is the layer consumers never think about and pay the most for when it’s done badly.

Why the next generation is harder to build

Smart devices are getting more capable, and that capability has a cost that lands squarely on the engineering team. More on-device intelligence. Stricter privacy rules. Longer battery expectations. Tighter security. Regulatory scrutiny that used to apply only to medical gear now creeping toward consumer products too.

None of this shows up in the marketing. The ad shows a person tapping a screen and a light turning on. What it doesn’t show is the year of board revisions, firmware rewrites, connectivity tests, and security audits that made that tap reliable.

So the next time a smart device just works, give a small nod to the invisible stack underneath. The clean experience on the surface is the product of a lot of unglamorous engineering refusing to cut corners. That refusal is the whole difference between a gadget you trust and one you return.

#Smart #Devices #Built #Engineers #Viewengineering,smart devices">How Smart Devices Are Actually Built: An Engineer’s View Pick up any smart device you own. A doorbell that recognizes faces, a watch that reads your heart rhythm, a thermostat that learns when you leave for work. They feel simple. You tap, they respond. That simplicity is a lie. A useful one, but a lie. Behind the clean app and the satisfying click is a stack of engineering decisions that most people never see. And the gap between a device that works for five years and one that dies in eight months almost always traces back to those invisible choices. So let’s look at what actually goes into building the connected gadgets shipping in 2026. Smart starts with the circuit board, not the cloud Most coverage of smart devices jumps straight to AI features and voice assistants. But the foundation is physical. A device is a printed circuit board, a microcontroller, a fistful of sensors, a radio, and a battery, all crammed into a shell that has to survive being dropped, sat on, and left in a hot car. This is where hardware development does its quiet, unglamorous work. Engineers pick a microcontroller based on how much computing the device needs versus how little power it can afford to burn. They route signal traces on the board so a Wi-Fi radio doesn’t drown out a delicate sensor reading. They run the whole thing through thermal testing, drop testing, and certification for FCC and CE marks before it can legally ship. Get this layer wrong, and no amount of clever software saves you. A poorly designed board produces flaky sensor data. Bad antenna placement means the device drops off your network the moment you walk to the next room. These aren’t software bugs. You can’t patch your way out of a physics problem. The companies building good hardware treat the proof-of-concept stage as a real checkpoint. They wire up development boards and modular parts to test the core idea cheaply, before committing to a custom design that costs real money to manufacture. It’s the boring discipline that separates products from expensive paperweights. Firmware is where the device actually thinks Sitting on top of the hardware is firmware. This is the low-level code that tells the chip what to do, when to wake up, how to read a sensor, and when to phone home. People mix up firmware and software all the time, so here’s the clean split. Software runs on your phone or in the cloud and handles the screens you tap. Firmware lives inside the device and controls the hardware directly. Firmware is genuinely hard to write well. The constraints are brutal. A typical IoT microcontroller has a tiny amount of memory, often measured in kilobytes, and it might run on a coin cell that needs to last a year. Every line of code competes for space and power. Then there’s timing. A lot of devices need deterministic, real-time behavior, meaning a sensor reading has to be processed within a fixed window or the whole thing falls apart. A heart monitor that processes a beat “eventually” is useless. The firmware has to guarantee it happens now. If you want the deep version of how this gets built in practice, Yalantis published a solid breakdown of firmware development for embedded IoT devices that covers architecture, power management, and the over-the-air update workflows that keep a device current after it ships. The OTA piece matters more than it sounds. A device that can’t safely update its own firmware is frozen in time the day it leaves the factory. Connectivity is a series of trade-offs Your smart device has to talk to something. Your phone, your router, a cloud server, or all three. Choosing how it talks is one of the most consequential engineering calls in the whole project, and there’s no single right answer. Bluetooth Low Energy sips power and works great for a wearable talking to your phone, but its range is short and it can’t reach the internet on its own. Wi-Fi reaches everything but drains batteries fast. LoRaWAN travels for miles on almost no power, which is perfect for a soil sensor in a field, but it carries tiny amounts of data slowly. Cellular options like NB-IoT and LTE-M let a device work anywhere there’s a signal, with the catch of ongoing data costs and bigger power draw. Engineers usually mix these. A fitness band might use BLE to sync with your phone, and your phone carries the data the rest of the way. An industrial sensor in a remote location might use LoRaWAN to a gateway, which then forwards everything over cellular. The “right” combination depends entirely on power budget, data volume, range, and cost, which is exactly why this decision gets made early and gets revisited often. Sensors and the messy job of trusting them A smart device is only as good as the data it collects. And raw sensor data is messy. Take a simple temperature reading. The sensor drifts over time. It gets warmed by the heat of the chip sitting next to it. It returns noisy values that jitter up and down even when nothing changes. Firmware has to calibrate, filter, and sanity-check all of it before the device acts on a single number. This gets serious fast in regulated fields. A continuous glucose monitor or a medical wearable can’t ship a reading that’s “close enough.” The sensor design, the calibration, and the firmware that validates the data all have to meet standards that consumer gadgets never face. The engineering bar is much higher, and the cost of getting it wrong is measured in patient safety, not customer reviews. For everyday devices the stakes are lower, but the principle holds. Good devices spend a lot of hidden effort turning unreliable physical signals into numbers you can actually trust. Where the AI hype meets the silicon Here’s the part that has changed most recently. A growing share of smart devices now run machine learning models directly on the chip instead of sending everything to the cloud. This is edge computing, and it’s reshaping how devices get built. The appeal is obvious. Processing data on the device means lower latency, since you’re not waiting on a round trip to a server. It means better privacy, because your data never leaves your hand. And it means the device keeps working when your internet goes down. The catch is that running a model on a chip with kilobytes of memory is an engineering puzzle. Models have to be shrunk, quantized, and optimized until they fit in the space available without melting the battery. The face-recognition that runs locally on a modern doorbell is a heavily compressed version of what would run on a server. Squeezing it down to fit is real, specialized work, and it’s increasingly where the competitive difference between two similar gadgets actually lives. Security can’t be the last step For years, connected devices treated security as an afterthought. Ship the product, patch problems later. That approach has aged badly. Outdated firmware is now one of the most common ways attackers break into IoT systems. Research from the security firm ONEKEY found that vulnerable firmware accounts for a large majority of successful attacks on connected devices. Once an attacker is inside one poorly secured gadget on your network, they have a foothold to reach everything else. Building security in from the start means encrypting data both when it’s stored on the device and when it travels to the cloud. It means signing firmware updates so a device only accepts legitimate code, not something an attacker swapped in. And it means designing for recovery, so a compromised device can be safely reset and restored rather than turned into a permanent liability sitting on your shelf. This is the layer consumers never think about and pay the most for when it’s done badly. Why the next generation is harder to build Smart devices are getting more capable, and that capability has a cost that lands squarely on the engineering team. More on-device intelligence. Stricter privacy rules. Longer battery expectations. Tighter security. Regulatory scrutiny that used to apply only to medical gear now creeping toward consumer products too. None of this shows up in the marketing. The ad shows a person tapping a screen and a light turning on. What it doesn’t show is the year of board revisions, firmware rewrites, connectivity tests, and security audits that made that tap reliable. So the next time a smart device just works, give a small nod to the invisible stack underneath. The clean experience on the surface is the product of a lot of unglamorous engineering refusing to cut corners. That refusal is the whole difference between a gadget you trust and one you return. #Smart #Devices #Built #Engineers #Viewengineering,smart devices

hardware development does its quiet, unglamorous work. Engineers pick a microcontroller based on how much computing the device needs versus how little power it can afford to burn. They route signal traces on the board so a Wi-Fi radio doesn’t drown out a delicate sensor reading. They run the whole thing through thermal testing, drop testing, and certification for FCC and CE marks before it can legally ship.

Get this layer wrong, and no amount of clever software saves you. A poorly designed board produces flaky sensor data. Bad antenna placement means the device drops off your network the moment you walk to the next room. These aren’t software bugs. You can’t patch your way out of a physics problem.

The companies building good hardware treat the proof-of-concept stage as a real checkpoint. They wire up development boards and modular parts to test the core idea cheaply, before committing to a custom design that costs real money to manufacture. It’s the boring discipline that separates products from expensive paperweights.

Firmware is where the device actually thinks

Sitting on top of the hardware is firmware. This is the low-level code that tells the chip what to do, when to wake up, how to read a sensor, and when to phone home. People mix up firmware and software all the time, so here’s the clean split. Software runs on your phone or in the cloud and handles the screens you tap. Firmware lives inside the device and controls the hardware directly.

Firmware is genuinely hard to write well. The constraints are brutal. A typical IoT microcontroller has a tiny amount of memory, often measured in kilobytes, and it might run on a coin cell that needs to last a year. Every line of code competes for space and power.

Then there’s timing. A lot of devices need deterministic, real-time behavior, meaning a sensor reading has to be processed within a fixed window or the whole thing falls apart. A heart monitor that processes a beat “eventually” is useless. The firmware has to guarantee it happens now.

If you want the deep version of how this gets built in practice, Yalantis published a solid breakdown of firmware development for embedded IoT devices that covers architecture, power management, and the over-the-air update workflows that keep a device current after it ships. The OTA piece matters more than it sounds. A device that can’t safely update its own firmware is frozen in time the day it leaves the factory.

Connectivity is a series of trade-offs

Your smart device has to talk to something. Your phone, your router, a cloud server, or all three. Choosing how it talks is one of the most consequential engineering calls in the whole project, and there’s no single right answer.

Bluetooth Low Energy sips power and works great for a wearable talking to your phone, but its range is short and it can’t reach the internet on its own. Wi-Fi reaches everything but drains batteries fast. LoRaWAN travels for miles on almost no power, which is perfect for a soil sensor in a field, but it carries tiny amounts of data slowly. Cellular options like NB-IoT and LTE-M let a device work anywhere there’s a signal, with the catch of ongoing data costs and bigger power draw.

Engineers usually mix these. A fitness band might use BLE to sync with your phone, and your phone carries the data the rest of the way. An industrial sensor in a remote location might use LoRaWAN to a gateway, which then forwards everything over cellular. The “right” combination depends entirely on power budget, data volume, range, and cost, which is exactly why this decision gets made early and gets revisited often.

Sensors and the messy job of trusting them

A smart device is only as good as the data it collects. And raw sensor data is messy.

Take a simple temperature reading. The sensor drifts over time. It gets warmed by the heat of the chip sitting next to it. It returns noisy values that jitter up and down even when nothing changes. Firmware has to calibrate, filter, and sanity-check all of it before the device acts on a single number.

This gets serious fast in regulated fields. A continuous glucose monitor or a medical wearable can’t ship a reading that’s “close enough.” The sensor design, the calibration, and the firmware that validates the data all have to meet standards that consumer gadgets never face. The engineering bar is much higher, and the cost of getting it wrong is measured in patient safety, not customer reviews.

For everyday devices the stakes are lower, but the principle holds. Good devices spend a lot of hidden effort turning unreliable physical signals into numbers you can actually trust.

Where the AI hype meets the silicon

Here’s the part that has changed most recently. A growing share of smart devices now run machine learning models directly on the chip instead of sending everything to the cloud. This is edge computing, and it’s reshaping how devices get built.

The appeal is obvious. Processing data on the device means lower latency, since you’re not waiting on a round trip to a server. It means better privacy, because your data never leaves your hand. And it means the device keeps working when your internet goes down.

The catch is that running a model on a chip with kilobytes of memory is an engineering puzzle. Models have to be shrunk, quantized, and optimized until they fit in the space available without melting the battery. The face-recognition that runs locally on a modern doorbell is a heavily compressed version of what would run on a server. Squeezing it down to fit is real, specialized work, and it’s increasingly where the competitive difference between two similar gadgets actually lives.

Security can’t be the last step

For years, connected devices treated security as an afterthought. Ship the product, patch problems later. That approach has aged badly.

Outdated firmware is now one of the most common ways attackers break into IoT systems. Research from the security firm ONEKEY found that vulnerable firmware accounts for a large majority of successful attacks on connected devices. Once an attacker is inside one poorly secured gadget on your network, they have a foothold to reach everything else.

Building security in from the start means encrypting data both when it’s stored on the device and when it travels to the cloud. It means signing firmware updates so a device only accepts legitimate code, not something an attacker swapped in. And it means designing for recovery, so a compromised device can be safely reset and restored rather than turned into a permanent liability sitting on your shelf.

This is the layer consumers never think about and pay the most for when it’s done badly.

Why the next generation is harder to build

Smart devices are getting more capable, and that capability has a cost that lands squarely on the engineering team. More on-device intelligence. Stricter privacy rules. Longer battery expectations. Tighter security. Regulatory scrutiny that used to apply only to medical gear now creeping toward consumer products too.

None of this shows up in the marketing. The ad shows a person tapping a screen and a light turning on. What it doesn’t show is the year of board revisions, firmware rewrites, connectivity tests, and security audits that made that tap reliable.

So the next time a smart device just works, give a small nod to the invisible stack underneath. The clean experience on the surface is the product of a lot of unglamorous engineering refusing to cut corners. That refusal is the whole difference between a gadget you trust and one you return.

#Smart #Devices #Built #Engineers #Viewengineering,smart devices">How Smart Devices Are Actually Built: An Engineer’s View

Pick up any smart device you own. A doorbell that recognizes faces, a watch that reads your heart rhythm, a thermostat that learns when you leave for work. They feel simple. You tap, they respond.

That simplicity is a lie. A useful one, but a lie.

Behind the clean app and the satisfying click is a stack of engineering decisions that most people never see. And the gap between a device that works for five years and one that dies in eight months almost always traces back to those invisible choices. So let’s look at what actually goes into building the connected gadgets shipping in 2026.

Smart starts with the circuit board, not the cloud

Most coverage of smart devices jumps straight to AI features and voice assistants. But the foundation is physical. A device is a printed circuit board, a microcontroller, a fistful of sensors, a radio, and a battery, all crammed into a shell that has to survive being dropped, sat on, and left in a hot car.

This is where hardware development does its quiet, unglamorous work. Engineers pick a microcontroller based on how much computing the device needs versus how little power it can afford to burn. They route signal traces on the board so a Wi-Fi radio doesn’t drown out a delicate sensor reading. They run the whole thing through thermal testing, drop testing, and certification for FCC and CE marks before it can legally ship.

Get this layer wrong, and no amount of clever software saves you. A poorly designed board produces flaky sensor data. Bad antenna placement means the device drops off your network the moment you walk to the next room. These aren’t software bugs. You can’t patch your way out of a physics problem.

The companies building good hardware treat the proof-of-concept stage as a real checkpoint. They wire up development boards and modular parts to test the core idea cheaply, before committing to a custom design that costs real money to manufacture. It’s the boring discipline that separates products from expensive paperweights.

Firmware is where the device actually thinks

Sitting on top of the hardware is firmware. This is the low-level code that tells the chip what to do, when to wake up, how to read a sensor, and when to phone home. People mix up firmware and software all the time, so here’s the clean split. Software runs on your phone or in the cloud and handles the screens you tap. Firmware lives inside the device and controls the hardware directly.

Firmware is genuinely hard to write well. The constraints are brutal. A typical IoT microcontroller has a tiny amount of memory, often measured in kilobytes, and it might run on a coin cell that needs to last a year. Every line of code competes for space and power.

Then there’s timing. A lot of devices need deterministic, real-time behavior, meaning a sensor reading has to be processed within a fixed window or the whole thing falls apart. A heart monitor that processes a beat “eventually” is useless. The firmware has to guarantee it happens now.

If you want the deep version of how this gets built in practice, Yalantis published a solid breakdown of firmware development for embedded IoT devices that covers architecture, power management, and the over-the-air update workflows that keep a device current after it ships. The OTA piece matters more than it sounds. A device that can’t safely update its own firmware is frozen in time the day it leaves the factory.

Connectivity is a series of trade-offs

Your smart device has to talk to something. Your phone, your router, a cloud server, or all three. Choosing how it talks is one of the most consequential engineering calls in the whole project, and there’s no single right answer.

Bluetooth Low Energy sips power and works great for a wearable talking to your phone, but its range is short and it can’t reach the internet on its own. Wi-Fi reaches everything but drains batteries fast. LoRaWAN travels for miles on almost no power, which is perfect for a soil sensor in a field, but it carries tiny amounts of data slowly. Cellular options like NB-IoT and LTE-M let a device work anywhere there’s a signal, with the catch of ongoing data costs and bigger power draw.

Engineers usually mix these. A fitness band might use BLE to sync with your phone, and your phone carries the data the rest of the way. An industrial sensor in a remote location might use LoRaWAN to a gateway, which then forwards everything over cellular. The “right” combination depends entirely on power budget, data volume, range, and cost, which is exactly why this decision gets made early and gets revisited often.

Sensors and the messy job of trusting them

A smart device is only as good as the data it collects. And raw sensor data is messy.

Take a simple temperature reading. The sensor drifts over time. It gets warmed by the heat of the chip sitting next to it. It returns noisy values that jitter up and down even when nothing changes. Firmware has to calibrate, filter, and sanity-check all of it before the device acts on a single number.

This gets serious fast in regulated fields. A continuous glucose monitor or a medical wearable can’t ship a reading that’s “close enough.” The sensor design, the calibration, and the firmware that validates the data all have to meet standards that consumer gadgets never face. The engineering bar is much higher, and the cost of getting it wrong is measured in patient safety, not customer reviews.

For everyday devices the stakes are lower, but the principle holds. Good devices spend a lot of hidden effort turning unreliable physical signals into numbers you can actually trust.

Where the AI hype meets the silicon

Here’s the part that has changed most recently. A growing share of smart devices now run machine learning models directly on the chip instead of sending everything to the cloud. This is edge computing, and it’s reshaping how devices get built.

The appeal is obvious. Processing data on the device means lower latency, since you’re not waiting on a round trip to a server. It means better privacy, because your data never leaves your hand. And it means the device keeps working when your internet goes down.

The catch is that running a model on a chip with kilobytes of memory is an engineering puzzle. Models have to be shrunk, quantized, and optimized until they fit in the space available without melting the battery. The face-recognition that runs locally on a modern doorbell is a heavily compressed version of what would run on a server. Squeezing it down to fit is real, specialized work, and it’s increasingly where the competitive difference between two similar gadgets actually lives.

Security can’t be the last step

For years, connected devices treated security as an afterthought. Ship the product, patch problems later. That approach has aged badly.

Outdated firmware is now one of the most common ways attackers break into IoT systems. Research from the security firm ONEKEY found that vulnerable firmware accounts for a large majority of successful attacks on connected devices. Once an attacker is inside one poorly secured gadget on your network, they have a foothold to reach everything else.

Building security in from the start means encrypting data both when it’s stored on the device and when it travels to the cloud. It means signing firmware updates so a device only accepts legitimate code, not something an attacker swapped in. And it means designing for recovery, so a compromised device can be safely reset and restored rather than turned into a permanent liability sitting on your shelf.

This is the layer consumers never think about and pay the most for when it’s done badly.

Why the next generation is harder to build

Smart devices are getting more capable, and that capability has a cost that lands squarely on the engineering team. More on-device intelligence. Stricter privacy rules. Longer battery expectations. Tighter security. Regulatory scrutiny that used to apply only to medical gear now creeping toward consumer products too.

None of this shows up in the marketing. The ad shows a person tapping a screen and a light turning on. What it doesn’t show is the year of board revisions, firmware rewrites, connectivity tests, and security audits that made that tap reliable.

So the next time a smart device just works, give a small nod to the invisible stack underneath. The clean experience on the surface is the product of a lot of unglamorous engineering refusing to cut corners. That refusal is the whole difference between a gadget you trust and one you return.

#Smart #Devices #Built #Engineers #Viewengineering,smart devices

Post Comment