The problem

NYT Cooking recipes live on a phone, and a phone is the worst tool in a kitchen: your hands are wet, floured, or holding a knife, and the screen locks every ninety seconds. Meanwhile the highest-intent moments in cooking happen away from any screen at all, staring at the bananas on the counter or standing in a store aisle deciding which pan to buy. Glasses put a camera and a display at both of those moments. No one at NYT was building for them.

What I built

A cooking assistant that lives on your face. Four connected flows:

Scan what you have. Look at an ingredient and ask about it. The demo question that sells it: "are the bananas still good?" The camera looks, the assistant answers.

Cook hands-free. Step-by-step recipes read to you and shown on the display, with voice control and countdown timers, so the recipe moves at your pace while your hands stay in the bowl.

Build the grocery list. Add items as you cook and as you scan, so the list writes itself while you notice what's missing or past its prime.

Shop with Gauge (Wirecutter). In the store aisle, get Gauge-tested product recommendations at the shelf. That one's the business story, not just the demo story.

Looking at the bananas: the vision flow

The banana question is small on purpose. It's the kind of question everyone asks in a kitchen and no one opens an app for, which is exactly why it works on glasses: the camera is already pointed at the fruit, so the interaction cost drops to a sentence. The prototype does camera-based ingredient recognition with a ripeness read, answers in voice, and can route the result somewhere useful, like adding replacement bananas to the grocery list.

The demo captures the personality too: a cooking assistant should sound like it likes food. That tone was a design decision, not an accident of the model.

Cooking at the speed of your hands

The recipe flow is built around one constraint: the cook never touches anything. Steps advance by voice. Timers get set by voice and count down on the display in your peripheral vision — you see the timer and the upcoming ingredients while your hands keep working. That's the whole pitch of the hardware in one interaction. A phone recipe interrupts cooking; this one keeps up with it.

The aisle moment

Hackathon projects usually skip the money question. I didn't want to. Wirecutter already does the work of testing products; the problem is that its recommendations live on a website and purchase decisions happen at a shelf. Kitchen closes that gap: standing in the aisle, wearing the glasses, you get the Wirecutter-tested pick for the thing in front of you.

That turns the assistant from a nice demo into a bundle story: Cooking supplies the recipes, the assistant supplies the context, and Wirecutter supplies a recommendation at the exact second of purchase intent. It's the most direct revenue line in the whole smart-glasses suite.

Why I built it

Self-initiated, no org mandate for wearables. I knew the hardware from prior work, and Maker Week 2026 was the opening. Kitchen was one of four glasses prototypes I built and deployed before the hackathon started. It's listed on the official Maker Week demos sheet.

What I'd tell another engineer

Pick a demo question before you pick an architecture. "Are the bananas still good?" decided everything downstream: it forced the camera flow, set the voice-answer format, and gave every stakeholder a one-sentence way to retell the project. If your prototype can't be asked for in one sentence by someone who saw it once, the build is fine and the framing is broken.