When you become a Snapdragon Insider, you expect early access to silicon, the occasional device, and a front-row seat to Qualcomm’s roadmap. What you don’t expect is a plane ticket to Mexico, an apron, and a camera crew waiting to watch you cook a dish you’ve never made before, with no recipe, no shopping list, and only on-device AI to guide you.
That’s exactly how I ended up here. As a French Snapdragon Insider who happens to spend as much time in the kitchen as I do chasing chip benchmarks, I was invited to put a very specific idea to the test: can artificial intelligence running entirely on your phone or your laptop turn a home cook loose on a foreign cuisine, in real time, with whatever happens to be in the fridge?
The brief: no recipe, no safety net
The rules were beautifully simple and slightly terrifying. I wasn’t told in advance what I’d be making. I opened the fridge, took stock of the ingredients in front of me, and from that, and only that, the AI helped me improvise a complete dish from a culinary tradition that wasn’t my own (and helped to translate some foreign words too).
No browsing through twenty recipe tabs. No carefully pre-measured mise en place. Just a French cook, a Mexican kitchen, an Indian dish, a fridge full of unknowns, and a Snapdragon-powered device acting as sous-chef. The whole point was to remove the script and see whether the technology could actually keep up with the chaos of real cooking.
Letting the AI lead
What surprised me most wasn’t that it worked, but how it worked. I described what I had, the AI proposed a direction, and from there it became a genuine back-and-forth: suggesting substitutions when I was missing something, adjusting quantities, walking me through techniques I wasn’t familiar with, and keeping me on track when I improvised a little too freely.
It stopped feeling like using a gadget and started feeling like cooking alongside someone who simply knew more recipes than I ever could.
And because the heavy lifting happened on the device rather than in the cloud (or both if needed), the experience was instant. No buffering wheel between “what do I do with this?” and an answer, which matters enormously when you’ve got a pan heating up and a clock running.
Why on-device AI actually matters here
I’ve written plenty on this blog about the Hexagon NPU and Qualcomm’s push to run generative AI directly on the phone instead of shipping everything off to a data center. This was the first time I felt that argument in my hands rather than read it on a spec sheet.
On-device AI means it’s fast, it works even when connectivity is patchy, and what you ask it stays with you. A kitchen is the perfect stress test: messy, time-sensitive, unpredictable, and deeply personal. If AI can be genuinely useful there, the “real-world AI” promise stops being a slogan.
The takeaway
I came to Mexico as a French cook expecting to be humbled, and left convinced that the most exciting use of AI isn’t writing emails or generating wallpapers, it’s the quiet, practical moment where it helps a real person do something they didn’t think they could. A foreign recipe, improvised from a stranger’s fridge, plated and eaten. That’s the future I want from the chip in my pocket.
Merci to the Snapdragon Insider team for the invitation, and yes, the dish was actually good.Â