Case study · Personal project“Any jellyfish?” Researched, designed, built and shipped solo. A question asked hundreds of times a day at every ladder and slipway on the coast, turned into a native iOS app that answers it honestly.
DÓNAL O'MAHONY · PERSONAL PROJECT · RESEARCH, DESIGN, BUILD, SHIP
Designers are building now. I have enjoyed that shift at work more than anything in years, so I wanted to try it properly on something of my own.
Figma for thinking. Claude Code in the terminal for the logic, the data modelling and the parts I would previously have specced out. Lovable for the app. Xcode to ship it.
A few evenings rather than months, using the same research and design discipline I have led teams on at Contentful and Verizon Connect, just compressed to one person.
A year ago I could not have built this.
The research and design discipline underneath it is the same one I have led teams on at Contentful and Verizon Connect, just compressed to one person and a few evenings instead of a team and a quarter.
Answer one question on one screen: should I get in? And be honest about how confident the answer is.
Carry live swimmer sightings, the thing every existing jellyfish map is missing, with no account and two taps.
Predict where nobody is measuring, in the eight months of the year when official bathing water sampling stops.
Prove that a designer with a clear vision can research, model, build and ship a native app alone.
Native iOS. Free. No account.
It started at home. My family are wary of jellyfish, in the sea here and every summer on holiday. Can we swim? Are there any in? What happens if one of us gets stung? Nobody could ever answer.
Then I noticed the same thing playing out near where I live, in public. At the ladders and slipways along the coast, there is a queue. People waiting to climb down into the water, and people climbing back out. And every time, without fail:
“Any jellyfish?”
The answer exists. It is in the head of the person climbing out, and it disappears the moment they reach their car.
Plenty of jellyfish maps exist. Marine institutes, citizen science databases, bloom trackers. Excellent for researchers. Almost none carry live user sightings. That is the gap. Someone on a ladder does not want a species distribution map. They want to know what the last person saw.
The answer was already at the water’s edge.
Each report can be confirmed or cleared in one tap. No account required.
Family, friends, the local swimming crowd. Conversations at slipways and in WhatsApp groups. The scope tripled, and every expansion came from them.
“I’d swim with jellyfish. I’m not swimming after heavy rain.”
Irish and UK coastal towns run combined sewers. Heavy rain overwhelms them and they discharge into the sea. Every regular swimmer knows this. Nobody knows when, because official notices follow fortnightly sampling and arrive days late.
That reframed the product: not where are the jellyfish, but is it safe to get in?
“I check three apps every morning.”
The highest frequency ask, and the least glamorous. The habit already existed. I just had to be the app they opened first.
One clear view of the day, not three weather apps.
Tide direction, sea temperature, moon, daylight and the best swim window remain readable as one system.
Unexpected. People wanted to talk about jellyfish: warming seas, longer seasons, new species. Older swimmers had long memories and strong opinions. That turned a utility into something with a bit of soul: species guide, sting first aid, what is in season and why.
Recognition and first aid belong in the same place.
Plain severity labels make the guide useful at the moment someone needs it.
Four systems sit behind the one screen. Each one reads a set of signals, makes a judgement a swimmer can act on, and refuses to do certain things. The refusals matter as much as the answers.
which species are in season and how badly they sting · sea temperature against seasonal norms · wind direction measured against each beach’s own orientation · tide state and movement · live swimmer reports.
whether an offshore bloom is likely to reach this bay today. A beach facing the wrong way in the same wind has a different answer to its neighbour.
to let quiet reporting read as good news. Sightings can only ever push risk up.
live rainfall from official gauges · how saturated the ground already is from the previous 48 hours · each catchment’s runoff behaviour · the travel time from overflow to that specific beach.
not is it raining, but has this network taken on more than it can hold, and when does that arrive here, often hours before an official notice.
to outrank an authority. Council and EPA notices always win.
tide height and direction · sea temperature · wind · daylight, sunrise and moon phase.
the practical question: when today is actually good? Most people plan around high water, not low.
to invent a window when there isn’t one.
all three models, plus what it told you yesterday.
a strict priority order: official notice, then modelled spill, then a fresh sighting, then jellyfish risk. The title carries the verdict in a few words. Arrives at a time you choose.
to repeat itself, or to wake you to say everything is fine.
This is the part a designer does not get to build. Working with an agent changed that. I brought the domain knowledge, the research and the judgement about what a swimmer would actually trust, and got a collaborator for the data plumbing and the API work. Not code generation. Genuine argument about whether a fallback was honest, and whether a number could be stood over.
The spill model in the product, hours ahead of an official notice
Every hard call got resolved by one of these.
An empty feed once said “All clear.” It felt great. It was a lie, because nobody reporting does not mean nothing is there. It usually means nobody looked.
Nobody has flagged jellyfish here recently. That’s not the same as all clear.
Less satisfying. Honest. The line I would want to stand over if someone got hurt.
Every figure carries provenance: Forecast, Verified, or both. Official notices are visually distinct from predictions, and official always wins. We can raise an alarm. We can never talk one down.
Official notices are reactive by design. In Ireland the monitoring season runs June to mid-September, while sea swimming is now year round. That gap is the product: modelled likelihood, hours ahead, available in the eight months when nobody is testing at all.
Coordinates never leave the phone. Every request is keyed by beach ID.
Free weather APIs are often non-commercial only. Open data is not the same as free to use. Rebuilding around commercially licensed sources, Environment Agency gauges, Met Éireann and EPA Ireland, produced something better than what it replaced: real tipping-bucket gauges within 7 km of every modelled outfall, instead of a model grid.
A one-time €2.99 unlock. Deliberately not a subscription, because pay once was the strongest line on the page. Then I built it: purchases SDK, entitlements, restore across devices, store products, sandbox testing, legal flows. And the sting in the tail. The moment an app takes money, non-commercial data licences stop applying to it. A €2.99 unlock did not just add a paywall. It changed the legal basis of every feed in the app.
So I took it out. Free, no purchase, no ads.
The relief was immediate. For a project made because the problem was interesting, free was not the fallback. It was the right answer.
Reports land where the decision is made.
Sightings and official water information share one map, close to the beach and current conditions.
Lead with the forecast, show season totals, never display a zero.
Lose weight gradually, stay visible as history. Confirming restores them.
Sea swimming runs on group chats. Someone answers “are we on?” every morning. That is the real user.
Jellyfish are summer. Water quality is year round.
The best finding was a question, not an answer.
“Any jellyfish?” was already being asked constantly. I did not create demand. I noticed a habit.
Users tripled my scope and were right every time.
None of it was in my first draft.
Honesty was the hardest constraint and the best one.
Every choice that made the product feel less confident made it more trustworthy.
Free was easier, more fun and cheaper.
Taking the paywall out removed a whole category of work, and the running costs with it.
The build culture I have enjoyed at work holds up on a hobby project.
Not for everything, and not without a lot of reading. But for something born from a question shouted up a ladder, a designer with a clear vision can now just make the thing, and have a good time doing it.
Built with Lovable, Claude Code, Capacitor and Xcode. Data from the Environment Agency, Met Éireann, MET Norway, EPA Ireland, Open-Meteo, OpenStreetMap and Mapbox.
