
Hey, we're Jesse and Michael. We were sick of weather apps being wrong so we built a better one.
Today we're launching the next iteration of Precip. It’s anchored on our own minute-by-minute rainfall forecast which is the most accurate in the US.
It all started with a gap most people don't know exists: accurate weather history.
We first ran into it a decade ago, while building a farming data-science startup in YC.
Most weather providers just show you the last forecast and call it history.
We do it differently.
We use high-resolution radar to measure what actually fell.
The result is 60% better rainfall measurements than just using NOAA's HRRR analysis.
Today we're introducing the next evolution of Precip: our own minute-by-minute rainfall forecast. Instead of a false alarm about rain that never shows up, or getting drenched without warning, Precip tells you when the storm will hit your precise location, how intense it'll be, and how long it'll last, right on your lock screen.
And it's not just guessing. Our benchmarks show Precip's minute-by-minute forecast is the most accurate in the US by a large margin: 85% more accurate than Apple Weather and 52% more accurate than Google Weather.
Here's what we found doing the research: There are dozens of published "state of the art" models that couldn't beat simple, non-deep-learning approaches, even ones from journals like Nature. None of the "foundation model" approaches shipped by the big players worked either.
The reason is twofold: low-quality training data (this is why we had to start with the most accurate history) and the wrong model architecture for this specific problem.
So if you're like us and don't want to get caught in the rain, download Precip: the first truly exceptional weather app since we lost DarkSky. https://link.precip.ai/get-precip
And if you're building your own product and want the best weather data, check out our API where you can pull data directly or connect your tools via MCP. https://api-docs.precip.ai