Tidesense, a month on: what happens when you point AI at something that actually helps

Follows on from I built a beach safety forecast for Jersey in a weekend with an AI browser.

When you live on a small island, the sea is never far from your mind. I’ve got an eight-year-old who’s falling more in love with the beach every week, and I’ve spent thirty years working in digital. At some point those two things were always going to collide.

Last month I wrote about building a beach safety forecast for Jersey in a weekend, well, I’m still building it a month on.

This is the bit I didn’t get to. Why I created it, what’s changed since, and what a month of real feedback has taught me.

What inspired Tidesense

AI gets a rough ride at the moment. Some of that’s fair. Between the billionaire tech bros and the fear-mongering, it’s easy to forget there are real positives in there too.

The reality is, used well, this stuff can do genuine good. We’ve been doing work with our friends at Scryptic and University College London, on a platform that teaches health and bioscience professionals the data science and machine learning behind modern research. Seeing that up close got me thinking. If this kind of technology can help people working on health and disease, why not use the same idea to help people get the most out of the ocean, and do it safely?

That’s where Tidesense came from. Not a product plan. Just a want to take the technology we’ve got, and the data that’s already out there, and point it at something useful.

For 25 years, Devstars has built digital work for some serious names. Creative work for the likes of Radiohead and Taylor Swift back in the day, then a queue-timing system for Heathrow Airport, a security platform for G4S, and a Ministry of Defence research programme into immersive training. I don’t mention that to show off. I mention it because, after all of that, the most satisfying thing you can build is sometimes the simplest one. The one that helps the people right around you.

What it does now

Quick recap, because it’s moved on since the last post.

Tidesense pulls in live data from a few places and blends it into one picture:

• Weather
• Tide times and heights
• Swell
• The actual shape of the island

That last one matters more than you’d think. Jersey’s cliffs and bays create wind shadows, spots that feel calm because the land is sheltering you. They also bounce wind back off the cliffs in ways that catch people out. So I’ve mapped the island’s topography in, to model where you’ll get that shelter and where you’ll get the bounce-back.

The water quality bit I’m most pleased with

Jersey checks its beaches through the summer for water safety and pollution. Generally our water is really good. But like anywhere with farmland nearby, heavy rainfall can wash run-off into the sea afterwards.

So I’ve built in Jersey’s own water quality data, and added an extra layer on top. If there’s heavy rainfall over a certain amount, Tidesense flags the day after as one to be more careful around the water. You wouldn’t want to be drinking seawater at the best of times. After a proper downpour, it’s just worth being a bit more switched on.

Where the data comes from

A forecast is only as good as what feeds it, so it’s worth being open about where Tidesense gets its information. I’m also just glad to credit the people who publish this stuff openly, because the app wouldn’t exist without them.

• Open-Meteo does the heavy lifting on weather and the sea. Wind speed, direction and gusts, temperature, rainfall, and the swell data behind the surf and bodyboard guidance. It also gives me the tide times. It’s free, well looked after, and accurate enough that I dropped the alternatives. (Side note for the tide nerds. The UK’s official Admiralty tidal service doesn’t actually cover the Channel Islands, so we’re a bit on our own out here. Open-Meteo comes within about twelve minutes and twenty centimetres of the official St Helier predictions, which is plenty good enough for planning a paddle.)
• Copernicus Marine Service, the EU’s ocean-monitoring programme, provides the currents. This is the science-grade stuff that professional oceanographers use, modelling how the water actually moves around the island hour by hour. It’s the main input behind the rip and drift warnings, which is the part I most wanted to get right.
• The States of Jersey (gov.je) publish the official bathing water sampling for our beaches. That sits behind the water quality verdict, and the rainfall rule I mentioned for St Aubin’s Bay. Nothing else carries the same weight, because it’s the island’s own testing.
• NOAA, the US ocean and atmosphere agency, runs a real-time wave buoy out in the South West Approaches. I use its live readings as a sanity check against the swell forecast, so I can show what the sea is actually doing, not just what the model predicts.
• OpenStreetMap contributors provide the maps behind the beach picker.

The thing that quietly pleases me about that list is how much of it is free, open, publicly funded data. People and organisations choosing to publish good information for anyone to use. Tidesense is really just a way of stitching it together for one job. Helping people use the ocean here, safely.

Data is one thing. Local knowledge is another.

You can model wind, tide and swell all day long. But the bloke who’s been launching off the same slipway for forty years knows things the API never will.

So I’ve built feedback into the app, so people can tell me where it’s right and where it’s wrong. That’s how it gets properly tuned. The plan is to fold that local knowledge back into the model over time.

If you know these waters, that’s a genuine ask. Tell me what it’s getting wrong.

Simple on the surface, detailed underneath

I’ll be straight with you. Right now the app is quite technical.

If you just want the headline, there’s a simple red, amber, green system. Green, crack on. Amber, be cautious. Red, not today. Most people can get what they need from that in a couple of seconds.

If you want to know why, the data’s all there, with overviews to help you read it. Over time I want to make the whole thing simpler for everyone, not just the data nerds. That’s the direction of travel.

The technical change that’s the same problem we solve for clients

One bit worth mentioning, because we sort this exact thing for businesses every week.

Tidesense started life as a JavaScript app. Great for people, not so great for search engines, which often struggle to read that kind of content. So I’ve built HTML layers over the top, the part search engines and AI tools can actually pick up and understand.

Behind those, I’m running PHP triggers and hourly cron jobs. Basically little automated jobs that wake up every hour, grab the latest data, and update the pages. So the public pages stay current, and you can click straight through to the live app from there.

It’s a small-island version of a problem we fix all the time. You can build something brilliant, but if search engines and AI can’t read it, half the world never finds it.

Still early, still honest about it

This is very much a side project. I’ve been building it on my own so far, more or less. The team have seen it, but I haven’t really pulled them in yet. That’ll come.

For now it’s beta, it’s honest about being beta, and it’s getting better every week, partly thanks to people telling me what’s wrong with it. It even got picked up by a sailing magazine recently, which I genuinely didn’t expect.

If you’re on the island and you love the water, give it a go. Tidesense is at tidesense.co.uk. Use it, break it, and tell me what local knowledge I’m missing, at stuart@devstars.co.uk.

And the thing that matters more than anything I’ve built. Always check RNLI safety advice, lifeguard flags, and the States of Jersey sea safety guidance before you get in. Tidesense is an early-stage, model-based forecast. It’s for guidance only, not a substitute for your own judgement or official safety advice.

---

Stuart Watkins is the founder of Devstars, a digital growth agency working from Jersey and London. He spends most of his time helping ambitious businesses build digital things that actually work, and the rest of it on the beach with his son.