The Tech Lowdown
Listen, we've all been waiting for AI glasses that actually do something useful without needing to send all our data to the cloud. PrismML, a lab founded by Caltech researchers, just dropped a major W for local processing. They’ve successfully shrunk their 'Bonsai' LLM to run directly on Qualcomm’s Snapdragon AR1 Gen 1 platform.
At the Snapdragon Summit this week, they showed off this 2-billion-parameter model. It’s tuned for vision and language, which means, theoretically, your smart glasses could identify whatever you're looking at in real-time. The best part? It’s not just a bulky model shoved into a small space—they’ve managed to compress it by 4x while keeping the performance high.
Why it matters
Real talk: the industry is currently obsessed with proprietary AI models that lowkey act like privacy vampires. PrismML is building for open-weight AI that lives on your hardware. If they can make this work, it’s giving a future where your device handles the heavy lifting locally, keeping your data off a random server.
It’s giving main character energy for edge computing, but don't get too hyped just yet—there aren't actually any consumer glasses running this tech available to buy. Say less, though; we're keeping an eye on this.
