Analysis

Frame is a pair of bluetooth smart glasses containing an ultra low power heads up display. It is made by Brilliant Labs and is part of a small wave of AI wearables that gained significant media coverage as part of the early 2024 AI hype wave. Brilliant is a small team (6) that effectively mobilized the latent desire many tech-people have for a completely minimalist, glasses-form-factor device. It is an execution-is-everything effort, compared to a deep tech and IP play. The primary innovation at play is a single round optomechanical display assembly that enables lower cost of production and a sort-of invisible style. The product launch went well gaining ~500k visits to their site in Februrary. From looking at discord order numbers, it seems they have sold at least 5000 pairs of $349 Frame glasses (~$1.5M), though they have had to walk back offering prescription lenses which may have changed the count. This makes the impact roughly similar to a successful Kickstarter launch, but not quite the breakout scale of the Rabbit R1.

The vast majority of their marketing is centered around AI, even though that does not appear to be the team’s core competency or something their hardware can support. The current application acts as a router with requests being fulfilled by a mixture of 3rd-party AI services such as Perplexity or OpenAI. They are VC- backed, raising $6m across two early rounds and likely make modest profits on hardware with future plans to generate higher-margin MRR through companion app subscriptions. They do have plans for companion devices, like a control ring, which could prove to be interesting.

Background: I have followed head-mounted display products for many years, am a computer engineer, and work in AI. I perfectly fit the target audience for this product. I first took note of Brilliant from their Monocle product and had infrequently checked in on their Discord to see if there was a ‘whats next’ product. I saw the Frame glasses launch and purchased them on release day.

Specifications

• Body: Flexible plastic body and arms. Arms have moulded-in flat wiring to button batteries located in the temple tips. 7mm thick, flat polycarbonate lenses that are round except for a flat spot at the top where the display is adhered.

• Display: 640x400 OLED microdisplay (Sony ECX336CN) with at 20º field of view at the top of the right lens. This is projected downwards reflecting on the bottom of the glasses before hitting the prism in the center of the lens and being directed into your eye.

  • Displayed as just text this means you have 25 by 9.7 characters to work with (the last line gets slightly clipped). Shown below.

    

  • What the display looks like when displaying the above X’s.

    

    
    

  • Currently the display only outputs text and pre-defined emoji, though there is a update supposedly coming that will enable drawing vectors and bitmaps.

  • The display supports showing only 16 colors at any one time.

• Compute: nrf52840 SoC processor which sits in the bridge of the glasses.

  • 64mhz, BTLE 5.3, 1 MB Flash, 256 KB RAM

  • FPGA (graphics and imaging)

• Camera: 1280x720 camera (captures 720x720)

• Microphone: single mic set on the inside left of the bridge

• Power: 210mAh built-in rechargeable li-ion battery

• Sensing: IMU for 6 degrees of freedom, software defined tap detection

• Software: Lua-based OS with a custom API

(they do not have any audio output)

Initial impressions

Upon getting the unit the first thing to note is the minimalistic packaging, which includes the device in a felt sleeve, a small nosepiece charger, and 2 bridge risers adapters. There is no manual or directions, just a small QR code on the outside of the package to download the companion app. The Glasses are smaller and lighter than I was expecting; even smaller than my normal glasses. There is no charge indicator on the glasses, nor any suggestion on the package on how to set the glasses up. In order to connect Frame to the app, you have to first connect the charger to the glasses and use a pin to ‘un-pair’ (reset) the device. After resetting, the app will detect the Frame and start a firmware update (which feels a little unsafe as you have no idea on the charge state). After that finishes you can use the Frame through the Noa app, which is effectively a pipe that allows you to access LLM services. One tap to wake, single tap to start listening (double tap to take a photo too), and another tap to send. In a bit more than normal GPT-4o latency an answer pops out on the screen. The service was able to answer some math problems I sketched out on a piece of paper, but about a quarter of the time I had it mention looking at a blurry image. That makes sense as the camera is positively microscopic, one of the smallest I have ever seen, so plan on sticking to well lit places and holding still. Beyond that there is, well nothing else to report as this is the only function of the glasses out of the box. The glasses have no UI, so do not expect ‘smartwatch for your face’ functionality. I personally don’t see myself using the Noa app, so I won’t dive into the details.

The display itself is pretty small and you quickly realize that it doesn’t actually show that much text at any given time. Pair that with no ability to scroll back and every interaction ends up feeling like watching subtitles, making it easy to miss something. It ends up feeling much closer to audio-only than expected.

I have had these for 2 weeks now and generally speaking: If you are buying this for any sort of consumer use (ie. with the Noa app exclusively), I am guessing you will be disappointed. If you are buying this to tinker and build apps, you will have some fun and some annoyances.

Notes: I skipped mentioning battery life as I never wore them long enough to really have that matter. It seems like they last at least a few hours with display always on. This should be good enough for most people. The glasses tend to discharge themselves even when on the charger, unless that is plugged into the wall.

The Frame itself

Positives

• The hackability. Open source everything (app, OS, firmware) is truly awesome and very unique. I really wish this was done by more young companies. It gives a great behind the scenes look, empowering backseat product developers to have opinions and make suggestions. (👋🏻)

• The form factor. Many people have been waiting to develop on a device like this (normal glasses form factor). These glasses are a good way of understanding the real feeling of HMDs well ahead of mass adoption.

• The simplicity. These are not complicated. There are only a handful of active components that make up the Frame glasses. I respect that. Not throwing the kitchen sink at a product like this takes guts.

  • The display seems to be fairly novel in the way that it reflects the image from a normal OLED without an large external prism.

  • It is inexpensive for a low-volume device.

• The quirks. The ‘Mister Power’ charger has a high smile factor initially.

• The fact it actually shipped. Most devices like this are vaporware. Brilliant announced the Frame in early Feb and I received it at the end of May. That is great!

Critiques

• The hardware. I have sympathy for how difficult it is to make polished end-user goods, but this really isn’t there yet in almost every meaningful way.

  • The display is incredibly finicky to get sharp. You have to place them with millimeter precision to really get the display to crisp up without vertical blur. This is because the glasses are too physically small. They sit high up, forcing me to slide them down my nose for a clear image – and I don’t have a particularly large face. Once sharp, the image seems exactly like it is coming from a small 640x400 reflected OLED display: bright in a dark room, but fades outdoors. I can see the pixels clearly. When you look at the display it feels like you are focusing quite near to your face which I do not find comfortable. This is not a media consumption device.

  • The lenses are thick, perfectly flat, and very reflective both on the main surface and the inner edges. This makes for a more annoying than expected wearing experience with glare a real issue (you can see behind you). Your right eye looks directly into the prism, meaning there are always these blurry horizontal lines in your vision even when the display is off. I doubt I could get used to this. In the dark the lens with the display has a fiber optic effect, glowing slightly. This means that you can see more reflections and the OLED display panel itself.

  • The plastic is thin and very flexible, what I would call wiggly. This makes them seem fragile in-hand and hard to put on with consistency. The nose pads are flat compared to other glasses I have, which makes them put more pressure on a small part of my nose — pretty uncomfortable. After 30 minutes I want to take them off.

  • The charger has been the bane of my usage so far, as the only way to pair the glasses requires poking a pin into a small reset hole on the charger, with the power cable and glasses attached. This absolutely needed to be a push button. It also does not shut the glasses off unless it is charged, which means that as soon as the small charger runs out of battery the glasses just stay on and drain power (quickly). So far they have been dead 90% of the time I have picked them up, and they take a few hours to charge. There is no charge status indicator, so you are always just guessing.

• The design. Cutting to the chase these will not be socially acceptable for 90% of people, they look pretty embarrassing. The design was created to be “unmistakable and distinctive” (their words) more than look clean. Thick, matte, round glasses with centered arms scream ‘I’m trying hard to be quirky!’. I get the business reasoning for standing out with a low-volume product, but IMO AR glasses should adopt a classic stance by default, quirky as an option.

  • I also understand that the round shape is needed for the display to function, but subtle changes still would have gone a long way here for appearance, adjustability, and comfort. Example:

    
    

  • I think the clear version would be my overall pick after having these in hand. They stand out no matter what, so might as well make it a conversation piece.

• The marketing. It is wild to call these ‘AI glasses’. There is absolutely no useful AI code that is running on-device. This is a developer-focused, bluetooth peripheral head mounded display. Period. We don’t call bluetooth earbuds ‘AI headphones’ do we?

  • The primary function of these glasses is I/O: IMU (motion), mic, and camera as inputs, with a display output.

  • Over hype and under deliver is just short-term thinking to ride a wave, which I dislike. Their intro video is blatantly misleading.

  • It is really weird that these were marketed as an end-user device when they so clearly are not ready for that yet.

    • The main text on the product page is: “Frame is designed to be worn as a pair of glasses with a suite of AI capabilities out of the box. Whether your daily pair of specs or workbench prototyping tool, Frame is ready for the journey.”

  • The Meta Ray Bans are likely 1000x more powerful and can really do AI on-device.

• The utility. This is an open section as with any development device there will be gains later.

  • The lack of a high resolution camera is really sad. I would have loved to see a big.LITTLE style camera arrangement used, even if the large camera was slow to transfer. A larger camera enables social use cases, long-term memory storage, facial recognition, and use cases where low-light/detail/legible text is a requirement (translation).

  • Bluetooth transfer only paired with no JPEG decoding on device is a near deal-breaker. Currently bitmap files need to be transferred which means you need to strongly degrade the bit depth or resolution in order to transfer the file in a reasonable amount of time (1-2fps). Hopefully the community can fix this quickly as it seems like a massive barrier to any real utility.

  • No native notifications on iOS is a big deal, which hurts many simple smartwatch use cases. Remains to be seen what can actually be done with Frame when your phone is locked and in-pocket.

  • The power vs. weight tradeoff was made too strongly. Processing power is extremely limited, 64mhz, and the device is very light at 39g. As a dev kit type device more compute and battery would be optimal even if weight went up 50%. Especially ephemeral local storage for things like fonts and graphics to make a nice launcher experience. Perhaps the similar Nordic nRF54H20 SoC can be used in the near future to enhance this without requiring much redesign.

    • This is important and annoying since it means dev effort will not transfer to future devices. Most of the ‘cool work’ with this headset will be in efficiently staying within the device’s constraints. That ends up being one-off custom work 90% of the time. A heavy now, light later compute platform like Snapdragon AR1 seems more future proof.

• The lack of forethought. This is a tiny team trying to solve both complex hardware and software problems. That puts them into crisis resolution mode 100% of the time meaning proactive work often gets postponed. This translates to numerous holes where they clearly never stopped to think from an end-user perspective. I got this device as one of the first few hundred, so this will improve in time, but as of now it sure is painful. Vital information is consistently missing; there is almost no direction when setting the device up, and there is no dedicated developer relations being done beyond chats in the discord group. This seems to miss the entire point of a developer kit device: to build up a strong, extensible open-source platform. Most people will have to re-invent the same paradigms just to get started. Trusting emergent open-source is just a default ‘meh’ option.

The potential

With that long list of negatives you may expect me to recommend no one buy them, but I think a certain type will have a lot of fun with these things. Currently there are no other devices that compete with Frame. I work in AI and have long been interested in helping push forward how people interact with technology; making i/o more seamless and higher throughput. Until Meta unlocks their next release to hacking, Frame provides the most accessible capture+feedback loop on the market. The strong constraints almost immediately let you understand which applications will never be useful on-face, how distracting a push notif will be, and the unavoidable limitations of current display hardware.

I have no doubt that these can be used to hype demos of all sorts of different LLMs, knowledge tools, and whatever else comes next. Accessibility features such as full-time subtitles, lip reading, and information overlay are specifically interesting and will push this device to its limits. Being open-source allows for extra optionality here, especially for developers to test out interaction patterns with local models and high-privacy applications. This makes an interesting companion to Open Interpreter or other computer-controlling project.

Conclusion

I have an unusual brain, so I was most interested to hack on Frame to demo memory augmentation capabilities. For this I would need to actually wear Frame to have it capture passive visual and audio data throughout my day. I find the glasses pretty uncomfortable after ~30mins, so that is a deal breaker. For now I will skip the Frame for this; the pin/pendant/earbud form factor is better. To use them for content/demo creation I would need to be able to record the display from a user perspective, something that is actually much harder than it might seem. For now I am toying with OS design and the creation of custom GPTs that can do all the coding, given the device has a very small solution space. Long-term I haven’t yet decided if I will keep them or let a new owner skip the current waitlist.

All devices like this have trade-offs and seeking perfection is often the reason products never make it into consumer hands. That said, I feel like Brilliant is playing an odd game here. They are marketing and building the Noa app for a wide audience to capture the AI moment, while delivering developer hardware that isn’t going to live up to that potential. This split focus leaves gaps in both approaches, putting Frame squarely in the luxury tinkerer realm. I am not new to early hardware or software and I had purposely set very low expectations for this device; I think it has just barely met that bar. None of this is a knock on the team, I want startups to succeed and I fully commend them for delivering such a complex product launch with so few resources. I look forward to rapid iteration on the software side and hopefully a much more powerful follow-on version.