Please... let's be serious right now. friend.com has been making the rounds recently, an AI wearable pendant; it doesn't even seem to be an assistant, just a piece of technology that is always listening to you and will send you a push notification to your mobile phone (only iOS at this time)... that's it.
And they're charging you $99 now for a pre-order, only available in the US and Canada.
So, you're considering pre-ordering this product? Even after seeing the Humane Pin and the Rabbit R1? What's the purpose or benefit that justifies this decision?
The company wishes to enter the hardware space, but the form factor cannot allow for any computation to happen on the device. This is evidenced by the fact that it requires a phone not just for setup but also for operation. This means it's just a fancy microphone with lights and haptics—don't forget the USB-C charging!
From a technical standpoint, this is a company attempting to break into a low-margin industry (hardware) by leveraging a high-cost service (GPT wrapper). But here's the kicker-there's no monthly subscription. This begs the question of how this product plans to survive beyond a few months. It's a red flag that potential buyers should not ignore.
My prediction is that either the product is never released or the value to the company or its investors comes from the "always listening" aspect of the product. We know that our phones and smart speakers forever listen in on us. Still, sadly for advertisers, companies such as Google, Amazon, Apple and "Phone manufacturers" aren't interested in giving that access to everyone else. You can't just install an app onto your phone to spy without the users questioning why that orange dot is on their status bar. So now we need a new way of spying and listening into what you are doing and where. And that's where I believe products like these will come in.
I get it. They claim to protect our privacy: "No audio or transcripts are stored past your friend’s context window." However, as of yesterday, when their terms and conditions were updated, they did not include anything to cover such promises. Which is strange (or not, since the product almost definitely doesn't exist, so why write legal documentation for it?).
They will create and release a relatively limited number of products to keep costs (not just manufacturing, QA and Customer Service but also powering whatever AI service they're running, for speech-to-text, speaker identification and generating a short response... if it's not just all handled by ChatGPT or similar with relatively simple prompt engineering) low and will just quietly fade into the background just like the other AI hardware technology have in the past.
Also, the timing is awful, besides the fact that they want to be listening in on you 24/7 even in the bathroom, why is this not a mobile app? At least that way you could deliver your own small model to the devices and have them all run on the native hardware, All phones released in the last year have NPUs and honestly, no one that isn't interested in having an NPU in their phone by Q1 2025 is going to care about your spy pendant.
Would I buy a wearable? What would my list of concerns be?
- Local computing - This matters because without local computing, this is just a piece of hardware making potentially expensive API requests for everything while also spying on me. Zero control, zero ownership, zero use without internet.
- Not easily covered/replaced by my phone—Since I already have a smartphone that knows me well and has seen and heard everything I have, why would I add another piece of tech to carry everywhere? How will this wearable help me remember things without my calendar? Will it assist me with interacting with others without access to my messages?
- Not just an LLM—Please, oh please, oh please... we are tired! The majority of things done or considered by these tools are worse than just googling. I don't want more hallucinations; I want accurate data retrieval and categorisation that can then be expressed in a human-readable format. I do NOT want an autocomplete algorithm to simply try and calculate what an answer to my question is. This is a Limitation to LLMs as we see them right now.
I'm not going to continue that list because I don't imagine myself, who is already deep in the Apple ecosystem, picking up an assistant tool that isn't also part of that ecosystem. Why would I trust a brand new company that has come out of nowhere with access to my emails, messages, contacts, notes and to listen in to my phone calls to be useful as an assistant? It's not reasonable in my eyes.