Telltale

About Telltale

Telltale shows you what the wind, tide and current are doing now and what they’ll do next, plans your route through them, and afterwards shows you how you actually sailed.

It runs on iPhone, Android, Apple Watch, Garmin watches, the web, Mac and Windows, and it’s built in Vancouver by Arek Sredzki, a sailor and physical AI engineer.

Record your next sail from your phone, your watch or a computer on the boat.

At a glance

Live conditions
Public stations and buoys, plus your own boat’s instruments
Forecasts
Our own models, trained on years of global forecasts and sensor data, updated in real time
Routing
Passage and race routes planned around the forecast, the tide and charted depths
Race analysis
Every tack measured against your polars, alongside the course and the fleet
Runs on
iPhone, Android, Apple Watch, Garmin, the web, Mac and Windows
Built by
Arek Sredzki, with 20 years of sailing and software experience

What Telltale does

Forecasting

Local knowledge, built into the forecast

Telltale trains its own models on years of global weather forecasts and real sensor measurements from stations, buoys and boats. They correct each new model run with what the sensors are reporting right now, producing high-resolution nowcasts and forecasts that update in real time and are built to beat the raw models where you sail.

Even the best weather models see the coast in grid squares kilometres wide. They miss the headland that bends the breeze, the inlet that funnels it and the forest that blocks it. Telltale reads the shape of the land and the seabed, learns how the wind and current behave around them from years of real sails and station records, and pulls each forecast toward what stations, buoys and boats are measuring right now.

Telltale archives every forecast and reading, and checks each model against what actually happened, station by station. Its own forecasts are held to the same standard. New methods are tested on days they have never seen before they reach the app.

Telltale Conditions over lower Howe Sound: the wind field flowing between Gambier, Keats and Bowen islands, over the terrain and seabed that steer it.
Howe Sound, where the mountains decide the wind.

Built by a physical AI expert

I’m Arek Sredzki, and I’ve been sailing and writing software for 20 years. I grew up racing 29ers on the water around Vancouver and spent a few summers teaching sailing.

At the University of British Columbia I was the software lead for UBC Sailbot, the student team that builds autonomous sailboats. We won the International Robotic Sailing Regatta, and we launched Ada, which set the record at the time for the longest distance sailed autonomously across the Atlantic. Planning her course meant working out how to route a boat on her own polars through weather forecasts that kept changing while she sailed. It’s the same problem Telltale’s routing tackles today.

For the last decade I’ve worked at the forefront of autonomous driving and safety-critical robotics. At Tesla I trained the end-to-end neural networks that power Full Self-Driving (FSD), led and shipped Actually Smart Summon, and helped develop and ship fully driverless robotaxis.

Training AI that has to perform in the real world teaches you what matters: real-world data at scale, models that learn from it, and proof against reality before anything ships. Telltale’s forecasting and analysis are built the same way, so every sailor gets the local knowledge and debriefs that used to take a professional navigator and coach.

These days I skipper Off Piste, a Salona 41, and we race her regularly out of Vancouver. I also sail a WASZP, which keeps me close to the foiling and skiff fleets. We test new features against our own recorded races before they ship.

arek.io

Ada, UBC Sailbot’s autonomous sailboat, sailing at sunset under a sail covered in sponsor logos.
Ada, UBC Sailbot’s autonomous Atlantic boat, under sail.
Off Piste, sail number CAN 1995, running under a white spinnaker past Stanley Park.
Off Piste in the 2026 Fraser River Lightship Race. Photo: Matthew Stranaghan.
A red WASZP, sail number CAN 2091, rigged on its dolly at the water’s edge.
The WASZP, ready to launch on English Bay.

Engineering

The principles behind Telltale

Every number has a source
A reading carries its station and time. A forecast carries its model, run and valid time. When Telltale adjusts a forecast with live data, it says so.
Uncertainty stays visible
When the models disagree, Telltale shows the spread side by side. It doesn’t quietly pick one and present it as the answer.
Dropouts are never hidden
When a feed or an instrument goes quiet, Telltale marks its last value with its age instead of passing it off as current.
Analysis never replaces the data
Tacks, percent of polar and highlights are derived from the recording. The raw readings stay available underneath, and a gap stays a gap.

Read the full methodology

Have a question, or know a station I’m missing?

Send a note through Support. I read every message.

Open ConditionsRecord a sail