Brilliant Labs 发布 399 美元 Halo AI 眼镜，主打持久记忆与自适应游戏

Brilliant Labs is bringing persistent AI memory and adaptive gaming to its Halo smartglasses, positioning the $399 wearable as a way to ask questions, create small applications, and turn everyday experiences into an evolving game. The company’s October 27 announcement puts its Noa assistant and a game experience called Narrative at the center of that approach.

From an AI Assistant to a Persistent Game

Brilliant Labs groups the experience into three activities: Ask, Make, and Play. Ask centers on Noa, a conversational assistant intended to remember preferences and earlier interactions. Instead of requiring users to reconstruct the same context with every question, the company says Noa can carry it forward.

Make extends that idea into voice-created miniapps and workflows. A user describes an experience they want to build, and the platform is designed to turn that request into a small application that can be shared through the Halo community’s app store. The attraction is a shorter path from an idea to a tool used in the moment, without the wearer having to write code.

Play is where the announcement takes a more unusual direction. Narrative is described as an adaptive game that incorporates real-world actions, discoveries, and decisions into an ongoing experience. Rather than following a fixed sequence designed by a studio, the game is intended to change with the player.

That is the company’s proposed experience, rather than a performance result established by this announcement. The meaningful test will be whether persistent context produces enjoyable, coherent play over time. Remembering an interaction is only the first step; a game also needs progression, understandable consequences, and reasons to return.

The Hardware Behind Halo

The launch specifications include a full-color microOLED display, microphones, bone conduction speakers, Bluetooth 5.3, and a weight just over 40 grams. Brilliant Labs claims up to 14 hours of battery life; actual endurance will depend on use.

The company’s hardware documentation identifies an Alif Balletto B1 processor with an Arm Cortex-M55 CPU and an Ethos-U55 neural processing unit. The latter provides dedicated hardware for supported local AI workloads. A low-resolution 640-by-480 global-shutter color sensor gives the system visual input, while dual microphones and stereo bone conduction audio support interaction without a handheld screen.

This combination illustrates the engineering tradeoff behind ambient AI: a device worn on the face has limited room for batteries, computing, and heat dissipation. Local inference hardware can handle selected tasks efficiently, but its presence does not establish that the entire assistant runs offline.

Brilliant Labs describes Halo as privacy-first. Its launch release says visual images are processed on-device and discarded, framing the sensor as a source of context rather than a conventional photography tool.

These are different parts of the product’s data flow. Local processing on the glasses should not be read as a promise that every Noa feature operates offline or that no visual information can reach a cloud service. For a memory-based assistant, the relevant questions include which features are enabled, what permissions have been granted, and how retained information is deleted.

Halo’s developer appeal rests partly on the ability to work below the application layer. Brilliant Labs publishes a Halo firmware repository built around Zephyr RTOS, with a Lua runtime for on-device applications and Bluetooth firmware updates. Its developer resources also point to SDKs and documentation.

That gives developers concrete building blocks for experimenting with wearable interfaces, sensors, and applications. The repository includes components with different licensing terms, so the availability of source code should be distinguished from a claim that every dependency carries the same unrestricted license.

For the wider smartglasses market, this approach makes Halo interesting as an extensible platform. Voice-generated miniapps could make customization accessible to consumers, while published firmware gives technical users a deeper route to building their own experiences. The two layers address different audiences but serve the same goal: letting the product evolve beyond its original feature set.

Pricing and the Next Test for Wearable AI

Halo costs $399. Noa and miniapps are included with daily usage caps, and the product page distinguishes the free Standard assistant from paid conversational tiers. Monthly pricing is $9.99 for Noa Plus and $39.99 for Noa Pro, making the ongoing service level part of the buying decision.

Brilliant Labs also announced backing from Goodwater Capital, with participation from Brendan Iribe and other investors, without disclosing a funding amount in the release.

Halo brings together several directions in consumer AI: persistent memory, natural-language app creation, wearable computing, and adaptive entertainment. Its opportunity lies in making those capabilities useful enough to become part of daily life. Delivering that experience will require both dependable interactions and clear boundaries around the personal context that makes them possible.