No friends on powder days

Welcome to Product Cocktail, where the takes are as polarizing as a shot of Fernet—but the insights come together like a perfectly crafted daiquiri.

Thursday, October 22nd, 2026. 11:37 PM.

"Babe, wake up."

"...whaaat?" my wife groggily replied.

"New prototype just dropped."

[Shoves laptop into her face.]

I wish I were kidding when I admit that's literally what happened last night.

I handed Claude Code a one pager I drafted, spent a handful of minutes answering questions, then let it rip agentic coding loops for 15 minutes before it delivered magic.

My literal reaction was frantically typing "holy shit this is wild" into the chat, then seven minutes later "this is literally insane, unreal - i want this product."

The scene: A day trip to Okemo Resort in Ludlow, VT in 2009 with my childhood friend, Andrew—one of my first skiing experiences without "adult supervision."

I was sitting on the frozen, icy chairlift climbing to the top of the mountain and I looked over at Andrew and mused "...what if we had first-person shooter-style heads up displays (HUD) in our ski goggles?"

I'm sure the conversation quickly turned to Mech Warrior 2: Mercenaries or Roller Coaster Tycoon nostalgia, but the ski goggle HUD idea stuck with me for seventeen years.

When ski tracking apps came on the scene in the mid-2010s, I was emboldened. With Ski Tracks, I had data at my chilly fingertips—elevation, runs/chairs taken, vertical feet, and most importantly, my top speed for the day.

In the last few years, with the arrival of Slopes, I adopted the "ski goggle heads up display" as a go-to Product Sense case interview answer to "what's your favorite app and how would you improve it?"

While preparing for a product interview over the summer, I landed on the product direction that resulted in this prototype above.

I didn't think much of the Meta Glasses product drop at first. "Oh great, a new generation of Google Glassholes" I thought at the time.

I started to hear rumblings that... this might actually be a compelling product? Then I saw the Oakley Meta Vanguard and my ski gearhead inner voice went off "Holy shit, it's a freaking pair of AI glasses in the body of a pair of Pit Vipers!"

Spoiler alert: they don't have a display. But it still got me thinking... what if?

I decided to build a concept prototype for a Meta Oakley x Slopes mashup. (This is obviously what any product manager with extra Claude tokens laying around would do. Obviously.)

I started from the customer problem.

The current hack: point and pray — or pull off a glove and dig out your phone (and hope they're on your "Find My Friends" list.)

The solution: an idea I came up with during the interview prep above.

One small rule ended up doing a lot of work: if you lose Carolyn, the HUD doesn't assume her destination until it's sure. Once she's past the last fork, it tells you where she's going to end up ("Carolyn → base of Emerald 6 Express"). It only shows that card when there's one destination left, so it's never wrong.

Here's the real one pager, if you're that much of a nerd.

The most exciting part about this entire prototyping experience is that I showed up to Claude Code not realizing the extent of what was even possible.

I came in wondering if this was even a good idea, given my disappointing revelation that the Oakley Vanguard doesn't have a display. Claude's first answer was no, not as written. Then it handed me a better idea:

From there, I gave it two Google Drive links to GoPro footage from Whistler three seasons ago, a Slopes export (GPS / skiing telemetry data) from the same day, and a one pager on the product direction.

Aside from some light adjustments (make it look like ski goggles, use this audio file for the voice commands, etc.), that's it.

Claude Code solved real technical challenges on its own.

The GoPro clock ran 33 seconds fast, which it found by matching the camera's motion sensor to GPS speed, confirmed by a mid-run stop and the arrival at the lift base.

A GPS glitch claimed a 91.5 mph top speed after a 2.5-minute dropout, which Claude corrected to 45 mph (turns out I’m not Lindsey Vonn).

It patched in real run and lift names by matching GPS tracks to OpenStreetMap. The chevron points at a fork I actually took, and the "next run" data on the chairlift is the run we actually skied next.

Tracked Carolyn by her jacket color to keep her name tag on her. I didn't tag a single frame. It even noticed her jacket reads navy in the shade and bright blue in the sun, and adjusted for it.

It took me 31 minutes from dropping in the one pager draft (10:59 pm) to a working HUD page in browser (11:30 pm). The build itself only took 14 minutes.

Overall, the project took 300 tool calls and rendered 4200 video frames for the final exports.

The best part: it barely dented a Claude Pro plan. I burned ~8% of the week's limit and didn't come close to hitting my session cap.

The work ended up being cheap because the heavy work ran locally on my MacBook Pro - ffmpeg (an open-source video transcoding library), and headless Chrome rendering the overlay one frame at a time. The model never processed video, it wrote scripts, read numbers, and looked at small screenshots.

Essentially: it did most of the thinking during the build step.

And it doesn't have to stop at glasses. Some of the most interesting skiing content on YouTube at this point is basically a guy narrating POV footage of them hitting sick lines with minimal on-screen graphics.

When I saw the demo, I started thinking this could be the beginnings of a real post-production pipeline for GoPro footage. Tag the route, overlay speed, and run names, for me, or any content creator.

To go even further, what if you layered in slope pitch, like onX Backcountry's slope-angle layer (RIP FATMAP)? Maybe even a subtle gradient color coding. (I dipped my toe into this without even realizing — onX's resort maps are built on the same OpenStreetMap ski data I used here.)

This was a cool demo, sure, but at this point it's a hope and a prayer that Meta or someone releases sport sunglasses-with-display (or god willing, a legitimate ski goggle).

What's missing here is the hardware and platform, not the product thinking. The Vanguard has no display and Meta's developer toolkit doesn't let you access the display anyway.

Meta put the screen on the wrong glasses. The everyday pair got it, the sport pair—worn where you can't pull out a phone—didn't.

I want this product, yesterday: https://www.youtube.com/watch?v=thsF5KJkbu0.

Get cracking, Zuck.

Send me questions, feedback, and cocktail recipes:
[email protected].

_{Icons made by Icongeek26 from www.flaticon.com.}