Brilliant Labs Updates Frame (and Halo!) SDK Docs & Thoughts.

Lots of new energy hitting the scene for us following the Brilliant Labs’s announcement of updates SDK docs. I noticed something a few days ago when poking about the Frame 👓 SDK for the concierge PWA experiment we’ve been doing. So it’s neat that there is indeed an update to things, along with a shift from “project” to “platform” in tone.

Going to be interesting to see/read (feel?) what some of you come up with. For some of us who played with the Humane AiPin, such things are just kind sticking around in persistent desires.

Chewing on a few things from the MS Build Conference keynote (and surrounding convos), the “platforming” Brilliant Labs is doing is neat… and likely points to a few other shifts some of us might take…

…Halo having its own (smaller) onboard model feels really similar to a part of the larger vision Microsoft’s Project Solara speaks towards. Some of the other glasses wearables who are relying on the mobile to be the only processor, and even reliance on heavier CPU/GPU/SOC hardware seems to be fading. There’s going to be networked inference, but the local capabilities will likely take a considerable step forward. Maybe a bit “Ready Player One,” maybe a bit worse. But definitely a shift similar to what AJAX did for web development (Gmail and Google Maps kinda made that wave if memory serves).

I don’t think “vibe coding” will stick around as it is either. I do think there’s going to be some shape of “make this for me” for some, but there’s a branch where I can see a “do this for me” and a “filter this from me” which evolves from this. Not good, not bad, just… different. Almost as if “agentic” might branch into a few directions.

I do think that if there’s a shape of Noa where a multi-LLM approach (agentic and beyond) which matures quickly shows up more and more often. It might even be the case that for “smaller” or “more personal” setups, that Noa’s platform approach becomes something of the norm (with the “Apple-like” approach where it’s more turn key, and probably more leaning to EU’s regs).

That’s part of what’s been making this Frame 👓 experiment really fun. It’s gone from project to something more. It’s become something more focused because of how I’ve been able to leverage several LLMs to get to working code, while also focusing on what needs to exist, versus what’s wish-list. It’s not yet “productivity” but it’s also not far either. There’s a capability being grown here where adding senses to oneself needs a human in the loop. The tension between “let the machine do it” and “do it myself” is strong. But perhaps as a result of this tension, new product categories define themselves in ways we have barely imagined.