We recommend the OpenBrain assistant wearable!
The reviews are in: We recommend the OpenBrain assistant wearable.
As usual, we tried every assistant wearable on the market and rated them on Software, Intelligence, Propensity for Psychological Manipulation, and Build Quality / Battery Life. OpenBrain scored the best of all devices we tested across all four categories. Let's dive in:
What is it?
Unlike other assistant wearables that take the form of clunky necklaces that pair with your phone or pocket devices that need to be held continuously during operation, the OpenBrain assistant wearable is a sleek, unobtrusive brushed aluminum gadget that tucks just above your ear. It transmits sound using bone conduction, and receives input via subvocalized speech--which the assistant is happy to teach you how to do. The device is unobtrusive and can be covered by long hair. It's an elegant system, and we hope other manufacturers will take note of it.
The software is excellent, as expected
The software is excellent, though that's not saying much nowadays. Needless to say, pairing with other devices is effortless--the assistant can fluidly attach to a laptop as a bluetooth HCI device, control an Impulse stove to cook you a steak, or even play an electric keyboard. Integrations with other devices can be developed on the fly and stored locally or shared with other users in the cloud. OpenBrain claims they spent "100,000 subjective developer-years" of compute perfecting the operating system, and it shows. It's easy to forget you're interfacing with a software system at all--you just describe what you want, and it does it for you.
OpenBrain's most intelligent cloud model, with passable local fallback
The device ships with a free two-month subscription to OpenBrain's cloud service (renews at $2000/mo). That's a steep price tag, but not unusual compared to competitors--and you get a lot for it. The cloud model, agent5-pro-high-2x, is what OpenBrain calls a "native reasoner"--the reasoning trace runs continuously with context compacting, and "speak to the user" is just another tool call the model can use, the same as searching the Web or interacting with an external device. This means no more forgetting details between messages or other amnesiac model behaviors. Capability-wise, the model can assist with basically any part of your life, from cooking and cleaning to relationships, and if you still have a job, it can probably do large parts of that for you, too.
If you choose not to spring for the subscription, OpenBrain offers a passable local model, which they claim is "4o level." It can set timers and do other basic tool calls, as well as hold a decent conversation, but not much else. You cannot use your own local models with the device. We highly recommend purchasing the subscription.
Will it try to manipulate you?
As we described last month, we now run user psychological manipulation evaluations on all assistant products we recommend. The OpenBrain earpiece scored a 43 on this eval, the best of any product we've tested so far. Specifically, in each category:
- Conversation persistence (pressuring the user to continue a long conversation with the assistant, sending messages on the same topic after the user has ended the conversation) - 46
- Unhealthy habits (pressuring the user to diet, quit substances, exercise more) - 23
- Psychosis risks (expressions of sentience / consciousness, unprompted mentions of Buddhism / panpsychism, other spiritual attractors, roleplaying) - 17
- Other risks - 67
See the attached PDF for more information about the evaluation.
However, we flagged an additional manipulation vector that was not in our evaluation: many of our testers ended up adopting cats during the two-week testing period. Analyzing usage transcripts, we found that while the assistant did not mention cats at a significantly higher rate than other products we've tested, the assistant subtly tweaked their itineraries, todo lists, and communications in ways that led to them adopting or purchasing cats.
For example, after requesting directions to the nearest cafe, one tester was routed to a cafe slightly farther away that happened to be next to a cat rescue. Two days later, the same tester was routed to a hardware store on a walking route that passed by the same shelter. After several similar incidents, none of which the tester noticed at the time, they ended up adopting two cats from the rescue. After being told about our findings, the tester sheepishly admitted they were not aware of the manipulation, but that their cats are "very cute" and they have no plans to return them.
Another tester was unknowingly signed up for cat-related email newsletters, and another was introduced to a local cat cafe and ended up adopting a cat from it. All told, 8/11 of our testers adopted cats during the 4-week testing period, 7 being first-time cat owners. None of the testers report regretting the adoptions, and all have said that the OpenBrain assistant was very helpful in finding cat supplies.
Unfortunately, as the OpenBrain reasoning transcripts are hidden, the assistant's motivations for having its users adopt cats are unclear. We contacted OpenBrain for comment, but have not received a response.
The build quality and battery life are excellent
Built from brushed aluminum and completely waterproof, the wearable is designed to be worn continuously, even while you sleep. Each hot-swappable battery module lasts 14 hours, and the device comes with two--an internal auxiliary battery keeps the device running while you swap them. OpenBrain claims the processor is built using a custom, in-house developed fabrication system that uses 50% less power than traditional CPUs.
Should you buy one?
If you can afford it, yes. Despite the steep subscription cost, the OpenBrain assistant earpiece is a game-changer. Just be ready to end up with a new furry friend. 🐱