Snowboarding's Next Upgrade Isn't Gear. It's Feedback. Heygo Debuts AI Snowboard Coach
New York, USA, Nov. 12, 2026 (GLOBE NEWSWIRE) -- Snowboarding is taught by people, and the people are hard to get. A full day of private coaching runs $989 to $1,372 at Winter Park and $1,250 to $1,470 at Deer Valley,1 and even where it is cheap, good ones are booked weeks out. So most riders do what Alexander Hardie did: watch YouTube, go up the hill, come down with no way of knowing whether they got it right.
Hardie is 39, lives in Ontario, and rides five or six days in a good winter, below the American average of 8.6. Instruction was never the scarce part: there is more coaching on YouTube than anyone could watch in a lifetime. What is scarce is a view of himself. On the hill there is no mirror and no coach, and no moment to check whether the edge he felt was the edge he meant to set.
Heygo is a technology startup selling what it calls an AI snowboard coach, and it is betting that the gap between knowing and feeling is a product. The kit is two inertial sensors that clip to the boots, an app, and a voice. It measures how the board is edged, how the body is balanced and how turns are timed, says something short while the rider is on snow, and hands over the day's report after the lifts close.
What comes back is not just a report. The system works out which fault produced each turn, then proposes the correction that fits the person riding: the same symptom draws a different prescription for a first-season rider and someone twenty years into the same habit. It also remembers, so day three can build on day one.
There is no camera, and nothing to operate at thirty miles an hour. The hardware costs $289 and ships with three months of the Pro subscription; after that the rider still enjoys basic coaching for free, while full and unlimited coaching costs $139 a year.
What it actually does
The sensors are inertial measurement units on the boots, not the board, so they stay with the rider, not the rental. The model resolves the board's edge angle to within two degrees and reads rotation several hundred times a second; pressure is not measured directly but inferred by the algorithm from how the boot loads and unloads through the turn. That is the difference between a device that describes a turn afterwards and one that speaks during it. A rider at thirty miles an hour cannot absorb a paragraph, so the voice gives short guidance after a segment, while it can still be acted on: too much weight on the back foot, change earlier.
The day view sits at a different altitude: edge angle across the day, the turns ranked, one or two things to work on tomorrow. The company calls the system Heygo IQ and describes the goal as turning every run into a repetition instead of a lap.
The ground truth problem
Heygo's engineers started at the bottom of the chain. The sensors sit outside the boot; what a rider cares about is the board. Between them are a soft shell, a binding and a body's weight. Under a hard edge load a boot can rotate 30 degrees while the board beneath it turns ten or twelve. Read the sensor as the board and every number downstream is wrong, and the errors compound: a few degrees at the ankle become a mistaken edge-change call, then a mistaken turn shape, then a confident, incorrect sentence in the rider's ear.
So the team built a reference set across resorts in New Zealand, Japan, Canada and China, on hardpack, ice and powder, with independent instrumentation: a board-mounted sensor and RTK positioning. Every run was filmed with time-synced video and annotated by instructors. More than two million turn-level data records now sit in that set, each carrying snow condition, boot stiffness, binding setup, rider weight and turn boundaries. No fixed parameter could cover that, because it changes with every boot, binding and kind of snow.
What comes out is compensation learned rather than assumed, and it shows in the numbers the company publishes: the boot misstates the board's edge by more than 20 degrees; after compensation, the error falls inside two degrees. Because the mapping comes from physical signals rather than images, there is no camera and no ritual: the rider clips the sensors on and rides, while the model works out where they were mounted and re-anchors when a fall knocks them out of alignment. Above that sits a layered system: calibration, then models that score edging, rotation, pressure and rhythm, then a diagnosis layer, then an agent deciding what to say.
The gap it is aiming at
For forty years the equipment got better and the coaching did not. What never changed is how a rider learns: someone stands on the hill, watches, says the same three things, and charges by the hour. Heygo's claim is not that it is better gear but that it is a new layer of gear, the first in decades that is not about the board or the boot but about what neither can tell you: what you just did.
Snowboarding is the growing half of a flat sport: 10.6 million American riders, up 9% and higher in seven of the past eight seasons, against 13.2 million skiers going nowhere. 2 81% of them are under 35. What the market has not produced is a device for that population: one that captures posture, reads what the rider did turn by turn, and returns advice written for that rider. On a snowboard, that space is empty.
The supply has not kept up: non-certified instructors at Vail start around $20 an hour, and a full day at Deer Valley, which still bans snowboards, runs $1,250 to $1,470. The gap is not a lack of coaching but a lack of coaching that reaches the days a rider actually rides. One lesson covers one of them, and the next starts from zero because nobody wrote down what happened.
Why AI belongs on a mountain
The AI hardware conversation has spent two years arguing about form factors and has curdled. Pew finds 52% of Americans more concerned than excited, up from 37% in 2021. 3 Camera glasses are barred from courtrooms, military bases and British venues. What the backlash is about is not intelligence but permission: devices that ask bystanders to accept being recorded and households to fund a new line item.
Sport is an honest test of whether it works. A rider at thirty miles an hour, in freezing temperatures and gloves, cannot look at a screen or wait. The intelligence has to be computed from sensors, delivered as speech. Heygo's answer is to make the AI deliberately narrow: no camera, two sensors, a model that reads edge and balance, guidance timed to the turn. That is the least fashionable AI hardware, and the version that finds durable buyers: narrow, sensor-driven, attached to an hour the customer already pays for.
Where it came from
Heygo began with a mistake no video could correct. Edward Wu, the company's founder and CEO, learned the way most riders learn: alone, from videos, run after run. A video can show you a good turn, but it cannot tell you that your edge is set late, that your weight sits on the back foot, that the turn you have practiced four hundred times is the wrong one. He kept repeating it, which is how an error turns into technique, until his lower back hurt and a season ended with an injury instead of a breakthrough.
What it is not
The most important sentence in Heygo's pitch is a negative one. "We are not replacing instructors," the company says. The device does two things instead. When no instructor is on the hill, it tells the rider how the run went. When one is, it hands that coach something to point at: edge angle, pressure trace, turn shape, so a student can see what was felt and keep it after the lesson. It was built with certificated instructors from CASI, AASI, BASI and SBINZ, whose progressions it treats as the reference rather than invent its own.
The pricing follows that logic, though not perfectly. Heygo is not a one-time purchase: $289 for the sensors with three months of Pro included, then $139 a year. Even with the renewal counted in, hardware and a year of coaching come in under one day of private instruction at a destination resort. For Black Friday the device is 15% off, with a seven-day free trial and three months of Pro included; orders go through heygo.ai.
Two riders, and what they do not prove
Leo is the case the company points to first, the least polished rider in the file. A software engineer who could link turns but had no plan past them, he set carving as a goal and found the problem underneath: he knew the destination, not the route. The report handed him one instruction per run, and he rehearsed it on the four-minute lift ride: weight onto the new edge through the front knee, change edges earlier, then move the pressure back. He climbed from the seventies into the hundreds and past 110, with a maximum edge angle of 48 degrees.
Sophia arrived with the opposite problem: a diagnosis and no way to verify it. A rider working toward a JSBA instructor certificate, she spent a month stuck between a skidded turn and a carved one, and one sentence from an instructor unstuck her. What she wanted was confirmation, run by run, without asking a friend to film every descent. She got numbers instead of shaky video, with the chatter on her heelside edge broken into the faults underneath. On a good morning she went from the low 110s to 138. Her verdict was blunt: "The data is objective, and that I like."
Two riders are not a study: both are paying customers who bought it themselves, and both were already hunting for evidence of their own progress, the disposition such a product selects for. What they establish is narrow: the loop closes for a rider willing to follow one instruction at a time.
What it is betting on
The bet underneath the product is narrower than the conversation it sits inside. Coaching has always been an event: an hour that arrives rarely, costs a lot and leaves no record. Heygo's argument is not that it coaches better than the person on the hill, but that improvement can become a habit, every turn measured, every mistake named, every correction repeatable on the other days of the season.
If that holds, it is selling something the sport has not had: not another upgrade to equipment, but the feedback layer the equipment never carried.
Hardie still has not booked a lesson for himself. The lessons in his house went to a seven-year-old, and nobody has sold him the same thing for the twenty years he has put in. That is the range Heygo claims: from the beginner who needs a report of the ride and a clear instruction, to the rider twenty years in who wants the numbers a coach would otherwise have to guess at.
About Heygo
Heygo makes an AI snowboard coach: two inertial sensors that clip to the boots, a voice while the rider is on snow, and a report after the day. Founded in 2025 by Edward Wu, a former ByteDance growth lead and CASI-certificated instructor, the company draws its team from ByteDance, Tencent and DJI. Heygo IQ was built with certificated CASI, AASI, BASI and SBINZ instructors and trained on more than two million turn-level data records. Available in the US at $289 with three months of Pro, then $139 a year.