Our principles

Three rules we don't bend.

01

Honesty over confidence.

If our AI confidence is below 70%, we tell you. If a camera is offline, we tell you. If we don't know, we say so. We'd rather be useful sometimes than impressive never-correctly.

02

Sources over opinions.

Every data point in RipCo is sourced from a named institution: the Bureau of Meteorology, Surf Life Saving NSW, partner camera networks. We cite. We timestamp. We don't editorialise.

03

People over algorithms.

RipCo will never replace a lifeguard. The flags are still the rule. Our role is to help you decide whether to drive to the beach in the first place, not to override the people on the sand.

AI Methodology

How rip detection actually works.

Plain-English explanation of the model behind RipCo. We don't pretend it's magic. It's a vision model trained on labelled Australian beach footage, running inference on every connected camera frame.

01 · INGEST

Live camera frame

An HLS stream from Surfline, Coastalwatch or Swellnet. We sample one frame every 4 seconds for analysis.

02 · DETECT

Vision model inference

A YOLO-family model fine-tuned on 12,000+ hours of labelled Australian beach footage scans the frame for rip current signatures.

03 · CONFIDENCE

Confidence score

Each detection comes with a confidence percentage. Below 70% we don't display it as a confirmed rip; we surface it as "possible".

04 · OVERLAY

Render to user

We overlay bounding boxes on the live feed, with the confidence number visible. The user always sees the source data.

12k+
Hours of labelled training footage
94%
Average detection confidence on validation set
4s
Inference cadence per camera
0
User images stored (we don't keep frames)
Where our data comes from

Named sources, every time.

Every data point in RipCo carries the institution it came from and the time it was last updated.

BOM

Bureau of Meteorology

Air temperature, precipitation, wind speed and direction, marine warnings.

bom.gov.au →
SLS
NSW

Surf Life Saving NSW

Patrol schedules, lifeguard tower locations, official beach closures, hazard advisories.

surflifesaving.com.au →
OPM

Open-Meteo Marine

Swell height, period, direction. Wave model output blended with BOM coastal observations.

open-meteo.com →
SF

Surfline · Coastalwatch · Swellnet

HD live cameras. Streams licensed via partnership; we don't rebroadcast outside the app.

WW

WillyWeather

Tide schedules, sun and moon times, UV index, regional weather backup.

willyweather.com.au →
RIP

RipCo Vision Model

In-house, trained on Australian footage in partnership with consenting beach club observers.

Honesty section

Where RipCo is wrong, slow, or silent.

What RipCo cannot do

  • RipCo does not replace lifeguards. If a beach has flags, swim between them. If it has no patrol, we strongly recommend a different beach.
  • Our AI cannot see what a camera cannot see, at night, in heavy rain, fog, or when surf height obscures the channel, detection confidence falls. We show this honestly; we don't fudge.
  • Conditions change in seconds. The score you saw 20 minutes ago in your kitchen may not be the score when you reach the water.
  • Sensor data has lag. Water temp can be 2 hours stale. Wind is a 10-minute average. Swell is modelled, not measured at every beach.
  • Community-reported hazards are verified before they fan out as alerts, but verification takes minutes, not seconds. SLSNSW is always the primary source for active emergencies.

In an emergency, call the people who can actually help.

RipCo is a planning tool. The numbers below are not.

SLSA · Australian Beach Safety

Always swim between the red and yellow flags.

It's the simplest, most important rule on any Australian beach. RipCo exists to help you decide which flagged beach to drive to, and to nudge you toward the patrolled ones.

Get the app

Smarter decisions, honestly told.

Free on the App Store. The data that drives the app, in your pocket.