frontpage.
newsnewestaskshowjobs

Made with ♥ by @iamnishanth

Open Source @Github

fp.

Open in hackernews

Show HN: Meteosource – Hyper-local weather app with minute-level nowcasting

https://play.google.com/store/apps/details?id=com.meteosource.weather.forecast.local.radar&hl=en
2•Sikara•1h ago
Hi HN,

I’m one of the creators of Meteosource Weather App. We have just released an update to our weather app and I would love to get your feedback.

Most weather apps are simple wrappers around a single API. We wanted to see if we could actually improve accuracy by using multiple models.

Instead of relying on a single provider, we built an engine that pulls from a multi-model ensemble (combining GFS, ECMWF, HRRR, etc.). We then apply machine learning to post-process these outputs — essentially training models on historical observations to identify and correct the systematic biases of each underlying numerical model for specific locations.

Key Technical Features: - ML-driven Nowcasting: We use real-time radar data and neural networks to predict precipitation to the minute - Bias Correction: Our models learn from past errors to improve local accuracy - Hyper-local resolution: We downscale global models to provide data for any specific coordinate

The App: - Precise hourly forecasts and interactive radar - Activity planning based on custom weather conditions - Minute-cast notifications and beautiful animated maps (in the pro version)

Comments

Sikara•1h ago
You can try the app for free: https://play.google.com/store/apps/details?id=com.meteosourc...

I’m here to answer any question!

andrewrozumy•1h ago
The multi-model ensemble approach makes sense — single provider apps are only as good as that one API. Curious how much the ML bias correction actually improves accuracy in practice versus the raw ensemble average. Do you have any before/after metrics on that?
Sikara•1h ago
Hi, the exact improvement heavily depends on the specific location and the weather patterns at the time. Generally though, we see roughly a 10% improvement in day-ahead forecasts for core variables like temperature and cloudiness compared to the raw ensemble average. The difference is even bigger if you compare it to the standard single-model approach that most weather APIs use.
andrewrozumy•1h ago
That's a meaningful improvement — 10% on day-ahead forecasts is actually significant when you compound it across a week of planning. Makes sense that location variance would be high though; I'd imagine coastal vs inland vs mountainous locations behave very differently for your correction models.

A pixel-art office for your AI agents

https://agent-office.pixelandprocess.de/
1•pixelandprocess•1m ago•0 comments

Iran conflict turns shipping market into 'Wild West'

https://www.ft.com/content/69f0f8a4-2567-44f4-a7b7-6ce4c62ab0e2
1•alecco•2m ago•1 comments

Why Your OpenClaw's Browser Keeps Getting CAPTCHAs

https://bennhuang.com/posts/why-your-openclaws-browser-keeps-getting-captchas/
1•bennhuang•4m ago•0 comments

Increased Errors on Opus 4.6

https://status.claude.com/incidents/0dvq4gvy5f5f
1•zurfer•5m ago•0 comments

Ruby 4.0.2 Released

https://www.ruby-lang.org/en/news/2026/03/16/ruby-4-0-2-released/
2•uticus•6m ago•0 comments

Nvidia Kimodo: kinematic motion diffusion model trained on mocap data

https://research.nvidia.com/labs/sil/projects/kimodo/
1•FrojoS•6m ago•0 comments

Experiments show potatoes can survive in lunar soil, with a lot of help

https://www.science.org/content/article/can-potatoes-grow-moon
2•jnord•7m ago•0 comments

Things I still wouldn't delegate to AI

https://www.jampa.dev/p/things-i-still-wouldnt-delegate-to
1•jampa•7m ago•0 comments

Are developers trusting AI-generated code too much?

3•skypark0407•7m ago•0 comments

The voice AI assistant that turns your thoughts into LinkedIn posts

https://edgarwrites.ai
1•teemupp•7m ago•0 comments

Libertify

https://www.libertify.com/
2•steve_rosenblum•8m ago•1 comments

Leading with decision-driven data analytics

https://caperio.ai/?p=5434
1•teleforce•9m ago•0 comments

Tell HN: Android blocks users from granting permissions to some apps

2•basilikum•9m ago•0 comments

COBOL Is the Asbestos of Programming Languages

https://www.wired.com/story/cobol-is-the-asbestos-of-programming-languages/
1•jim_lawless•10m ago•1 comments

Virtual File System for Node.js: +18,945 LOC Leveraging Claude Code

https://github.com/nodejs/node/pull/61478
1•cdrnsf•11m ago•0 comments

7-ton asteroid causes fireball seen over Ohio

https://www.wlwt.com/article/loud-boom-ohio-meteor-cleveland-cincinnati/70768910
1•pseudolus•11m ago•0 comments

Data Organization in Spreadsheets (2018)

https://www.tandfonline.com/doi/full/10.1080/00031305.2017.1375989#abstract
1•Tomte•11m ago•0 comments

The Oscars Died This Weekend, Did Anyone Notice?

https://quoththeraven.substack.com/p/the-oscars-died-this-weekend-did
1•RickJWagner•12m ago•0 comments

Paul Ehrlich Wasn't Premature: He Was Wicked

https://www.realclearmarkets.com/articles/2026/03/18/paul_ehrlich_wasnt_premature_he_was_wicked_1...
2•RickJWagner•13m ago•0 comments

What do new nuclear reactors mean for waste?

https://www.technologyreview.com/2026/03/18/1134345/advanced-nuclear-reactors-waste/
1•Brajeshwar•13m ago•0 comments

Jailrun: Orchestration for isolated environments, powered by FreeBSD

https://jail.run/
1•zaiev•14m ago•0 comments

Show HN: Run Away From Work – a browser game built with React and TypeScript

https://nyaomaru-portfolio.vercel.app/game
1•nyaomaru•14m ago•0 comments

Aggregated File System (AGFS), a modern tribute to the spirit of Plan 9

https://github.com/c4pt0r/agfs
1•mpweiher•16m ago•0 comments

Show HN: Backproto – network backpressure routing applied to AI agent payments

https://www.backproto.io/
2•BSOhealth•17m ago•0 comments

Ask HN: Is Claude down Again?

4•rreyes1979•17m ago•4 comments

IDE Runs Your Code. Now It Runs Your AI

https://joshuadarron.medium.com/your-ide-already-runs-your-code-now-it-runs-your-ai-65acfde22f66
1•joshuadarron•17m ago•1 comments

Inference Price Index – what AI inference costs across 8 providers

https://www.volthq.dev/
1•JackArnot•18m ago•1 comments

Macquarium

https://en.wikipedia.org/wiki/Macquarium
1•arbesman•18m ago•0 comments

Show HN: We built a way to try 50 AI models from one API

1•sharur7•19m ago•0 comments

Crewless Russian fuel tanker adrift near Malta called "ticking time bomb"

https://www.ctvnews.ca/world/article/ticking-time-bomb-stricken-russian-fuel-tanker-drifts-near-t...
1•heresie-dabord•19m ago•0 comments